Compare commits

..

71 Commits

Author SHA1 Message Date
Slavi Pantaleev
f809bf8d7c Prepare 1.19.0 release
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-05-09 16:34:49 +03:00
Slavi Pantaleev
44aea427fc Update matrix-sdk to v0.17.0 and Rust toolchain to 1.95.0
matrix-sdk 0.17.0 brings a few breaking API changes that we adapt to:

- Drop the `native-tls` feature on matrix-sdk; it was removed upstream
  in 2026-04 (matrix-sdk now hardwires rustls). The previous workaround
  comment referencing rust-mxlink issue #1 is no longer relevant.
- `Relation::Reply` is now a tuple variant wrapping a `Reply` struct;
  pattern matches updated to bind through `reply.in_reply_to.event_id`.

Also bumps the Rust toolchain from 1.93.0 to 1.95.0 (closes #85, which
proposed the Dockerfile bump in isolation). rustc 1.94+ trips a
query-depth overflow when computing async layouts in the matrix-sdk
timeline future graph; matrix-rust-sdk PR #6489 raises the limit, but
\`recursion_limit\` is per-crate, so we repeat \`#![recursion_limit = "256"]\`
on this crate root.

Bumps the mxlink floor to 1.14.0, which carries the matching adapter.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-05-09 16:29:41 +03:00
renovate[bot]
8f45a57ced Update docker.io/ollama/ollama Docker tag to v0.23.2 2026-05-09 16:06:25 +03:00
renovate[bot]
6a9100309c Update ghcr.io/element-hq/synapse Docker tag to v1.152.1 2026-05-09 16:06:17 +03:00
renovate[bot]
1953878e6b Update Rust crate tokio to v1.52.3 2026-05-09 08:12:03 +03:00
renovate[bot]
35761a5bf7 Update ghcr.io/element-hq/element-web Docker tag to v1.12.17 2026-05-09 08:11:38 +03:00
renovate[bot]
c7fa0cc0ee Update forgejo.ellis.link/continuwuation/continuwuity Docker tag to v0.5.9 2026-05-09 08:11:25 +03:00
renovate[bot]
87fcf9d019 Update dependency prek to v0.3.13 2026-05-09 08:11:04 +03:00
renovate[bot]
4b52bc906c Update forgejo.ellis.link/continuwuation/continuwuity Docker tag to v0.5.8 2026-04-25 22:16:54 +03:00
renovate[bot]
517a6c5e33 Update docker.io/ollama/ollama Docker tag to v0.21.2 2026-04-24 08:17:02 +03:00
renovate[bot]
636c8a35eb Update Rust crate async-openai to 0.36.0 2026-04-24 06:54:26 +03:00
renovate[bot]
a50de600da Update docker.io/ollama/ollama Docker tag to v0.21.1 2026-04-22 07:15:30 +03:00
Slavi Pantaleev
e978d3cb2f Publish only after successful CI
Run the publish workflow from the workflow_run event so Docker publishing happens only after the CI workflow completes successfully for push events on main or v* tags.

Check out the exact SHA validated by CI and derive Docker metadata from the upstream CI ref, so publishing follows the tested revision instead of the default branch tip.
2026-04-20 22:09:37 +03:00
Slavi Pantaleev
2b1bdbd3d2 Split CI and publish workflows
This supersedes 7d183b9 ("Run CI for pull requests"), which mixed validation and publishing in one workflow and regressed docker-manifest by dropping the package-write permission it needs to publish the manifest.

Split the workflows so CI handles pull requests, branch pushes, tags, and manual runs, while publishing stays focused on Docker delivery with the manifest permission fixed explicitly at the job level.
2026-04-20 22:08:07 +03:00
Slavi Pantaleev
7d183b91d1 Run CI for pull requests 2026-04-20 21:40:03 +03:00
renovate[bot]
420f380417 Update Rust crate async-openai to 0.35.0 (#125)
Co-authored-by: renovate[bot] <29139614+renovate[bot]@users.noreply.github.com>
2026-04-20 21:39:09 +03:00
renovate[bot]
6f541e2361 Update forgejo.ellis.link/continuwuation/continuwuity Docker tag to v0.5.7 2026-04-17 21:52:46 +03:00
renovate[bot]
0737c2761e Update docker.io/ollama/ollama Docker tag to v0.21.0 2026-04-17 10:41:04 +03:00
renovate[bot]
45852a0d53 Update Rust crate tokio to v1.52.1 2026-04-17 10:38:22 +03:00
renovate[bot]
0fcf8d8703 Update Rust crate tokio to 1.52.* 2026-04-15 09:43:04 +03:00
renovate[bot]
41905b006a Update docker.io/ollama/ollama Docker tag to v0.20.7 2026-04-14 07:33:22 +03:00
renovate[bot]
ee3d27701b Update docker.io/ollama/ollama Docker tag to v0.20.6 2026-04-13 08:58:29 +03:00
Slavi Pantaleev
1b5c2fbdc1 Prepare 1.18.0 release
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-04-11 10:55:55 +03:00
Slavi Pantaleev
a865a26093 Update tiktoken-rs to 0.11, adding support for newer GPT models
Supersedes https://github.com/etkecc/baibot/pull/116

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-04-11 10:52:22 +03:00
Slavi Pantaleev
1563e83672 Update dependencies 2026-04-11 10:40:09 +03:00
renovate[bot]
cf114b37b3 Update docker.io/ollama/ollama Docker tag to v0.20.5 2026-04-10 09:21:42 +03:00
renovate[bot]
b8ef2b978b Update Rust crate tokio to v1.51.1 2026-04-08 19:25:58 +03:00
renovate[bot]
7e37aee1b9 Update ghcr.io/element-hq/synapse Docker tag to v1.151.0 2026-04-08 11:57:52 +03:00
renovate[bot]
53836e556a Update ghcr.io/element-hq/element-web Docker tag to v1.12.15 2026-04-08 11:57:29 +03:00
renovate[bot]
89052bbfd9 Update ghcr.io/element-hq/element-web Docker tag to v1.12.14 2026-04-07 19:53:11 +03:00
renovate[bot]
07eb12d406 Update docker.io/ollama/ollama Docker tag to v0.20.3 2026-04-07 08:45:11 +03:00
renovate[bot]
3272cd6fb2 Update Rust crate tokio to 1.51.* 2026-04-04 15:52:04 +03:00
renovate[bot]
7ab97a39ca Update docker.io/ollama/ollama Docker tag to v0.20.2 2026-04-04 10:35:45 +03:00
renovate[bot]
11c5a9942e Update docker.io/ollama/ollama Docker tag to v0.20.1 2026-04-04 09:17:02 +03:00
renovate[bot]
251dec454c Update docker.io/ollama/ollama Docker tag to v0.20.0 2026-04-03 07:40:18 +03:00
renovate[bot]
1031dbf672 Update docker.io/ollama/ollama Docker tag to v0.19.0 2026-03-30 08:31:02 +03:00
renovate[bot]
581f00b9fb Update docker.io/ollama/ollama Docker tag to v0.18.3 2026-03-26 08:18:33 +02:00
Slavi Pantaleev
e57778d2bd Prepare 1.17.0 release 2026-03-25 20:18:56 +02:00
kschwank
2d659964a7 Add sender context mode for text generation (#104)
Add a per-room/global `sender_context_mode` setting that optionally
prefixes conversation messages with sender metadata before sending them
to the model provider.

This helps models distinguish between participants in multi-user rooms.
                                                                                                                                                                                                                                                             
Three modes are supported:
- `disabled` (default, no change)
- `matrix_user_id` (prefixes with `[sender=@user:server]`)
- `matrix_user_id_and_timestamp` (adds `send_at`; Example: `[sender=@user:server sent_at=<ISO 8601>]`)
                                                                                                                                                                                                                                                             
Sender context is applied to user and assistant text messages only,
skipping system prompts and non-text content.

Mixed-sender merged turns (something we intentionally do for Anthropic)
have their `sender_id` cleared to avoid misattribution.
2026-03-25 20:14:34 +02:00
renovate[bot]
661e7263fb Update ghcr.io/element-hq/synapse Docker tag to v1.150.0 2026-03-24 17:08:21 +02:00
renovate[bot]
1705c16762 Update ghcr.io/element-hq/element-web Docker tag to v1.12.13 2026-03-24 14:28:10 +02:00
Slavi Pantaleev
2455117e41 Prepare 1.16.1 release 2026-03-24 14:05:35 +02:00
Slavi Pantaleev
3e9c110afc Update dependencies 2026-03-24 14:04:40 +02:00
Slavi Pantaleev
d9b5524c97 Fix OpenAI response input for async-openai 0.34
async-openai 0.34 adds a required phase field to EasyInputMessage,
which broke our Responses API request construction.

Set phase to None because baibot does not currently model assistant commentary vs final-answer turns,
so omitting phase preserves the previous behavior while keeping the request compatible with the new crate.
2026-03-24 12:51:56 +02:00
renovate[bot]
9b169a7d28 Update Rust crate async-openai to 0.34.0 2026-03-24 12:40:25 +02:00
Slavi Pantaleev
cb29419d75 Prepare 1.16.0 release
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-03-20 12:25:52 +02:00
Slavi Pantaleev
51ca8c9948 Update dependencies 2026-03-20 12:23:51 +02:00
Slavi Pantaleev
12b938d2d1 Skip files with application/octet-stream MIME type
Files with unrecognized MIME types (application/octet-stream) are
not supported by any LLM provider and would cause errors that
permanently break the conversation thread. Instead, represent them
as a text message describing the attachment so the LLM is still
aware a file was sent without the thread becoming unusable.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-03-20 12:22:38 +02:00
Slavi Pantaleev
f3d1b32ad7 Use mime_guess for file extension MIME type detection
Replace the hand-maintained extension-to-MIME mapping with the
mime_guess crate, which was already in the dependency tree.
This covers hundreds of file extensions out of the box.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-03-20 12:22:38 +02:00
Slavi Pantaleev
3d3bd3c9f9 Add support for file attachments (m.file) in conversations
Files sent as m.file Matrix messages are now downloaded, MIME-detected,
and forwarded to LLM providers alongside the conversation context,
similar to how m.image is already handled.

- OpenAI provider: sends files inline as base64 data URLs
- Anthropic provider: skips files with a warning (library limitation)
- OpenAI-compat provider: skips files with a warning (library limitation)

Controller routing respects the existing prefix requirement setting.
MIME detection expanded to cover PDF, text, code, and document formats.
Docs updated to reflect file support and known limitations.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-03-20 12:22:38 +02:00
renovate[bot]
a25845e89e Update Rust crate quick_cache to v0.6.21 2026-03-19 19:22:00 +02:00
renovate[bot]
479f54b93d Update docker.io/ollama/ollama Docker tag to v0.18.2 2026-03-19 08:15:14 +02:00
renovate[bot]
90ab6807ac Update Rust crate quick_cache to v0.6.20 2026-03-18 09:15:53 +02:00
renovate[bot]
5022f79bf5 Update docker.io/ollama/ollama Docker tag to v0.18.1 2026-03-17 07:25:56 +02:00
renovate[bot]
2ba3b5a437 Update docker.io/ollama/ollama Docker tag to v0.18.0 2026-03-14 22:34:33 +02:00
renovate[bot]
2819011b9d Update Rust crate tracing-subscriber to v0.3.23 2026-03-13 20:44:43 +02:00
renovate[bot]
ede9065f77 Update Rust crate async-openai to v0.33.1 2026-03-13 11:54:12 +02:00
renovate[bot]
5e61f5b3a3 Update ghcr.io/element-hq/synapse Docker tag to v1.149.1 2026-03-11 20:05:48 +02:00
renovate[bot]
8633e82f62 Update Rust crate quick_cache to v0.6.19 2026-03-11 20:05:32 +02:00
Slavi Pantaleev
6aa3d70d57 Bump default OpenAI text-generation model (gpt-5.2 -> gpt-5.4) 2026-03-11 09:53:02 +02:00
renovate[bot]
75e29ee3fb Update Rust crate tempfile to 3.27.* 2026-03-11 08:50:06 +02:00
renovate[bot]
eab7978b9f Update ghcr.io/element-hq/element-web Docker tag to v1.12.12 2026-03-10 22:06:39 +02:00
renovate[bot]
765e7c17b2 Update ghcr.io/element-hq/synapse Docker tag to v1.149.0 2026-03-10 20:19:06 +02:00
Slavi Pantaleev
748d2b7fd4 Prepare 1.15.0 release
Add the 1.15.0 changelog entry covering access-token authentication support, Rust 1.93.0 toolchain pinning, docs updates, and dependency updates. Bump crate version to 1.15.0 in Cargo.toml and Cargo.lock.
2026-03-07 11:13:42 +02:00
Slavi Pantaleev
527759dd02 Document Matrix authentication modes
Add a dedicated configuration/authentication doc covering password and access-token setup,
including environment-variable mappings and token-generation example.

Link to it from configuration docs and align the sample config wording with Matrix Authentication Service/OIDC terminology.
2026-03-07 11:08:50 +02:00
Slavi Pantaleev
3290255bad Pin local Rust toolchain to 1.93.0
Add rust-toolchain.toml so local development and ad-hoc cargo commands use the same known-good compiler version as CI.
This avoids the matrix-sdk query-depth overflow seen on newer stable toolchains.

Related to 9b987395b3

Ref: https://github.com/etkecc/baibot/pull/83#issuecomment-4008463792
2026-03-07 10:40:56 +02:00
Slavi Pantaleev
9b987395b3 Pin GitHub CI Rust toolchain to 1.93.0
Using the floating stable toolchain currently breaks this project. Repro via just build-debug:

```
Compiling matrix-sdk v0.16.0
error: queries overflow the depth limit!
help: consider increasing the recursion limit by adding #![recursion_limit = "256"] to your crate (matrix_sdk)
note: query depth increased by 130 when computing layout of matrix-sdk Client::sync async body
error: could not compile matrix-sdk (lib) due to 1 previous error
```

Pinning CI to 1.93.0 keeps CI on the known-good toolchain until upstream/toolchain compatibility is addressed.
We should have had that to begin with (we're pinning as much as we can anyway), but..

Ref: https://github.com/etkecc/baibot/pull/83#issuecomment-4008463792
2026-03-07 10:38:35 +02:00
Taylor Southwick
4852d1fe92 Add support for access tokens using MAS (#83)
* Add support for access tokens using MAS

* use 1.13.0

* Update dependencies

* Harden auth credential selection in matrix link init

Use the same non-empty access-token criterion for auth mode selection and bind the token directly from the branch condition.
Return explicit configuration errors for missing or empty `device_id`/`password` instead of panicking, so invalid auth config fails gracefully.

* Centralize and harden user auth config handling

Move authentication-mode resolution into typed config parsing with ConfigUserAuth,
so downstream login setup consumes validated credentials instead of re-checking raw optional fields.

Enforce explicit password-vs-token selection, validate token/device/user-id requirements in one place,
and normalize empty auth env overrides to unset values for consistent behavior across YAML and environment input.

* Add auth config unit tests

Move auth_config tests into a dedicated cfg test module file to keep production config code compact while preserving behavior coverage. The tests cover password/token mode selection, missing/both auth method rejection, missing device_id, and empty-value handling.

* Use conventional mxlink version requirement

Replace the unconventional wildcard lower-bound expression with a standard semver lower bound for readability and tooling consistency.

---------

Co-authored-by: Slavi Pantaleev <slavi@devture.com>
2026-03-07 10:26:40 +02:00
renovate[bot]
8bd313f0d4 Update docker/build-push-action action to v7 2026-03-06 10:15:37 +02:00
renovate[bot]
711e1099d6 Update docker.io/ollama/ollama Docker tag to v0.17.7 2026-03-06 08:16:12 +02:00
renovate[bot]
91c8dd8f7d Update docker/metadata-action action to v6 2026-03-05 22:22:11 +02:00
54 changed files with 1610 additions and 615 deletions

26
.github/workflows/ci.yml vendored Normal file
View File

@@ -0,0 +1,26 @@
name: CI
on:
workflow_dispatch:
pull_request:
branches: [ "main" ]
push:
branches:
- "**"
tags: [ "v*" ]
permissions:
contents: read
pull-requests: read
concurrency:
group: ci-${{ github.event.pull_request.number || github.ref }}
cancel-in-progress: true
jobs:
test-and-clippy:
name: Unit testing and linting
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v6
- uses: dtolnay/rust-toolchain@1.93.0
- name: Install SQLite3
run: sudo apt-get update && sudo apt-get install -y libsqlite3-dev
- run: cargo test --all-features
- run: cargo clippy

View File

@@ -1,44 +1,49 @@
name: CI (main and tags)
name: Publish
on:
push:
branches: [ "main" ]
tags: [ "v*" ]
workflow_run:
workflows: [ "CI" ]
types: [ "completed" ]
permissions:
checks: write
contents: write
packages: write
pull-requests: read
contents: read
concurrency:
group: ${{ github.workflow }}-${{ github.ref }}
group: publish-${{ github.event.workflow_run.id || github.ref }}
cancel-in-progress: false
jobs:
test-and-clippy:
name: Unit testing and linting
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v6
- uses: dtolnay/rust-toolchain@stable
- name: Install SQLite3
run: sudo apt-get update && sudo apt-get install -y libsqlite3-dev
- run: cargo test --all-features
- run: cargo clippy
docker-clean-metadata:
if: |
github.event.workflow_run.conclusion == 'success' &&
github.event.workflow_run.event == 'push' &&
(
github.event.workflow_run.head_branch == 'main' ||
startsWith(github.event.workflow_run.head_branch || '', 'v')
)
runs-on: ubuntu-latest
outputs:
json: ${{ steps.meta.outputs.json }}
steps:
- name: Checkout
uses: actions/checkout@v6
with:
ref: ${{ github.event.workflow_run.head_sha }}
fetch-depth: 0
- name: Extract metadata (tags, labels) for Docker
id: meta
uses: docker/metadata-action@v5
uses: docker/metadata-action@v6
with:
images: |
ghcr.io/${{ github.repository }}
tags: |
type=raw,value=latest,enable=${{ github.ref_name == 'main' }}
type=raw,value=latest,enable=${{ github.event.workflow_run.head_branch == 'main' }}
type=semver,pattern={{raw}}
docker-build:
if: |
github.event.workflow_run.conclusion == 'success' &&
github.event.workflow_run.event == 'push' &&
(
github.event.workflow_run.head_branch == 'main' ||
startsWith(github.event.workflow_run.head_branch || '', 'v')
)
permissions:
contents: read
packages: write
@@ -57,6 +62,9 @@ jobs:
steps:
- name: Checkout
uses: actions/checkout@v6
with:
ref: ${{ github.event.workflow_run.head_sha }}
fetch-depth: 0
- name: Log in to the GitHub Container registry
uses: docker/login-action@v4
with:
@@ -65,10 +73,10 @@ jobs:
password: ${{ secrets.GITHUB_TOKEN }}
- name: Extract metadata (tags, labels) for Docker
id: meta
uses: docker/metadata-action@v5
uses: docker/metadata-action@v6
with:
tags: |
type=raw,value=latest,enable=${{ github.ref_name == 'main' }}
type=raw,value=latest,enable=${{ github.event.workflow_run.head_branch == 'main' }}
type=semver,pattern={{raw}}
flavor: |
latest=auto
@@ -77,13 +85,23 @@ jobs:
ghcr.io/${{ github.repository }}
- name: Build and push Docker images
uses: docker/build-push-action@v6
uses: docker/build-push-action@v7
with:
push: true
tags: ${{ steps.meta.outputs.tags }}
labels: ${{ steps.meta.outputs.labels }}
docker-manifest:
if: |
github.event.workflow_run.conclusion == 'success' &&
github.event.workflow_run.event == 'push' &&
(
github.event.workflow_run.head_branch == 'main' ||
startsWith(github.event.workflow_run.head_branch || '', 'v')
)
permissions:
contents: read
packages: write
needs:
- docker-build
- docker-clean-metadata

View File

@@ -1,3 +1,52 @@
# (2026-05-09) Version 1.19.0
- (**Internal Improvement**) Update [matrix-sdk](https://crates.io/crates/matrix-sdk) from 0.16 to 0.17 and [mxlink](https://crates.io/crates/mxlink) to 1.14.0. matrix-sdk 0.17 dropped its `native-tls` feature and now uses [rustls](https://github.com/rustls/rustls) exclusively as its TLS backend.
- (**Internal Improvement**) Bump the pinned Rust toolchain from 1.93.0 to 1.95.0 (in `rust-toolchain.toml` and the Docker build images).
- (**Internal Improvement**) Dependency updates.
# (2026-04-11) Version 1.18.0
- (**Bugfix**) Fix the bot not sending a welcome message when joining a room on homeservers (like [Continuwuity](https://continuwuity.org/)) that place the join membership event in the sync response's `state` block rather than the `timeline` block, via [mxlink](https://crates.io/crates/mxlink) 1.13.1
- (**Improvement**) Update [tiktoken-rs](https://crates.io/crates/tiktoken-rs) to 0.11, adding tokenization support for newer GPT models (gpt-5.x, codex, etc.) and fixing context sizes for o1-mini/chatgpt-4o/gpt-4.5
- (**Internal Improvement**) Dependency updates
# (2026-03-25) Version 1.17.0
- (**Feature**) Add `text-generation sender-context-mode` for attaching sender metadata to conversation messages. See the [💬 Text Generation](./docs/configuration/text-generation.md#-sender-context-mode) documentation for details. Thanks to [kschwank](https://github.com/kschwank) for the contribution in [#104](https://github.com/etkecc/baibot/pull/104)!
# (2026-03-24) Version 1.16.1
- (**Bugfix**) Fix compatibility with [async-openai](https://crates.io/crates/async-openai) 0.34.0 by populating the new `phase` field required for OpenAI Responses API message inputs. baibot does not currently distinguish between assistant `commentary` and `final_answer` turns, so using `None` preserves the previous behavior while remaining compatible with the updated crate.
- (**Internal Improvement**) Dependency updates.
# (2026-03-20) Version 1.16.0
- (**Feature**) Add support for file attachments (`m.file` Matrix messages) in conversations. Files like PDFs, text documents, spreadsheets, code files, etc. are now downloaded and forwarded to the LLM alongside the conversation context, similar to how images (`m.image`) are already handled. See the [💬 Text Generation](./docs/features.md#-text-generation) documentation for details and known limitations.
- (**Improvement**) Use the [mime_guess](https://crates.io/crates/mime_guess) crate for MIME type detection from file extensions, replacing a hand-maintained mapping. This covers hundreds of file extensions out of the box.
# (2026-03-07) Version 1.15.0
- (**Feature**) Add support for authentication via access tokens (for [Matrix Authentication Service](https://github.com/element-hq/matrix-authentication-service)/OIDC-enabled homeservers) as an alternative to password authentication. See [🔐 Authentication](./docs/configuration/authentication.md) for setup details. Thanks to [Taylor Southwick](https://github.com/twsouthwick) for the contribution in [#83](https://github.com/etkecc/baibot/pull/83)!
- (**Internal Improvement**) Pin the Rust toolchain to `1.93.0` in both CI and local development to avoid `matrix-sdk` build failures on newer stable toolchains.
- (**Internal Improvement**) Documentation updates.
- (**Internal Improvement**) Dependency updates.
# (2026-02-18) Version 1.14.3
- (**Internal Improvement**) Add [Renovate](https://docs.renovatebot.com/) configuration for automated dependency updates

1041
Cargo.lock generated

File diff suppressed because it is too large Load Diff

View File

@@ -7,7 +7,7 @@ license = "AGPL-3.0-or-later"
readme = "README.md"
keywords = ["matrix", "chat", "bot", "AI", "LLM"]
include = ["/etc/assets/baibot-torso-768.png", "/src", "/README.md", "/CHANGELOG.md", "/LICENSE"]
version = "1.14.3"
version = "1.19.0"
edition = "2024"
[lib]
@@ -17,24 +17,23 @@ path = "src/lib.rs"
[dependencies]
anthropic = { git = "https://github.com/etkecc/anthropic-rs.git", branch = "fix-content-block-image" }
anyhow = "1.0.*"
async-openai = { version = "0.33.0", features = ["audio", "chat-completion", "image", "responses"] }
async-openai = { version = "0.36.0", features = ["audio", "chat-completion", "image", "responses"] }
base64 = "0.22.*"
chrono = { version = "0.4.*", default-features = false, features = ["std", "now"] }
# We'd rather not depend on this, but we cannot use the ruma-events EventContent macro without it.
# We add the `native-tls` feature, because of https://github.com/etkecc/rust-mxlink/issues/1
matrix-sdk = { version = "0.16.0", default-features = false, features = ["native-tls"] }
matrix-sdk = { version = "0.17.0", default-features = false }
mime_guess = "2.0.*"
mxidwc = "1.0.*"
mxlink = ">=1.13.0"
mxlink = ">=1.14.0"
etke_openai_api_rust = "0.1.*"
quick_cache = "0.6.*"
regex = "1.12.*"
serde = { version = "1.0.*", features = ["derive"], default-features = false }
serde_json = "1.0.*"
serde_yaml_ng = "0.10.*"
tempfile = "3.26.*"
tiktoken-rs = { version = "0.9.*", default-features = false }
tokio = { version = "1.50.*", features = ["rt", "rt-multi-thread", "macros"] }
tempfile = "3.27.*"
tiktoken-rs = { version = "0.11.*", default-features = false }
tokio = { version = "1.52.*", features = ["rt", "rt-multi-thread", "macros"] }
tracing = "0.1.*"
tracing-subscriber = { version = "0.3.*", features = ["env-filter"] }
url = "2.5.*"

View File

@@ -4,7 +4,7 @@
# #
#######################################
FROM docker.io/rust:1.93.1-slim-trixie AS build
FROM docker.io/rust:1.95.0-slim-trixie AS build
RUN apt-get update && apt-get install -y build-essential pkg-config libssl-dev libsqlite3-dev

View File

@@ -4,7 +4,7 @@
# #
#######################################
FROM docker.io/rust:1.93.1-slim-trixie AS build
FROM docker.io/rust:1.95.0-slim-trixie AS build
RUN apt-get update && apt-get install -y build-essential pkg-config libssl-dev libsqlite3-dev

View File

@@ -22,6 +22,8 @@ You can see the list of supported environment variables in the [🦀 src/entity/
> [!WARNING]
> The static configuration contains an `initial_global_config` key, which is used to populate the bot's global configuration (stored as [dynamic configuration](#dynamic-configuration)) the first time the bot starts. Modifying this subsequently will not have any effect. After initial global configuration creation, it's expected to be managed dynamically via chat commands.
For Matrix-account authentication setup, see [🔐 Authentication](./authentication.md).
### Dynamic configuration

View File

@@ -0,0 +1,23 @@
## 🔐 Authentication
baibot supports 2 authentication modes for the Matrix account (`user.*` keys in config).
Set **exactly one** mode. If both are set (or neither is set), startup validation fails.
### Password authentication
- Config key: `user.password`
- Environment variable: `BAIBOT_USER_PASSWORD`
### Access token authentication
- Config keys: `user.access_token` + `user.device_id`
- Environment variables: `BAIBOT_USER_ACCESS_TOKEN` + `BAIBOT_USER_DEVICE_ID`
Access-token authentication is useful for OIDC-enabled homeservers (e.g. those using [Matrix Authentication Service](https://github.com/element-hq/matrix-authentication-service)).
Example token-generation command:
```sh
mas-cli manage issue-compatibility-token <username> [device_id]
```

View File

@@ -8,7 +8,7 @@ You can also use **different models within the same room** (e.g. [💬 text-gene
The bot supports the following use-purposes:
- [💬 text-generation](../features.md#-text-generation): communicating with you via text (though certain models may "see" images as well)
- [💬 text-generation](../features.md#-text-generation): communicating with you via text (though certain models may also process images and files)
- [🦻 speech-to-text](../features.md#-speech-to-text): turning your voice messages into text
- [🗣️ text-to-speech](../features.md#️-text-to-speech): turning bot or users text messages into voice messages
- [🖌️ image-generation](../features.md#image-generation): generating images based on instructions

View File

@@ -57,6 +57,25 @@ This feature relies on [tokenization](https://en.wikipedia.org/wiki/Large_langua
This setting is **disabled by default**, but can be enabled via `!bai config room text-generation set-context-management-enabled true` (this can also be set globally, see [🛠️ Room Settings](./README.md#room-settings)).
### 👤 Sender Context Mode
In multi-user rooms, it may be useful for the model to know which participant sent each message in the conversation context.
To support this, the bot has a `text-generation sender-context-mode` setting, which can be set to:
- (default) `disabled`: do not attach sender metadata to messages before sending them to the model
- `matrix_user_id`: prefix text messages with the sender's Matrix user ID, for example: `[sender=@alice:example.com] Hello bot`
- `matrix_user_id_and_timestamp`: prefix text messages with the sender's Matrix user ID and the message timestamp, for example: `[sender=@alice:example.com sent_at=2026-03-23T14:30:00Z] Hello bot`
This sender metadata is attached to conversation messages before they are sent to the model provider. It applies to user and assistant text messages, but not to system prompts or non-text content.
⚠️ Enabling this sends Matrix user IDs, and optionally timestamps, to the model provider.
Example: `!bai config room text-generation set-sender-context-mode matrix_user_id` (this can also be set globally, see [🛠️ Room Settings](./README.md#room-settings))
### ⌨️ Prompt Override
You can override the [system prompt](https://huggingface.co/docs/transformers/en/tasks/prompting) configured at the [🤖 agent](../agents.md) level.

View File

@@ -8,7 +8,7 @@ You can also use **different models within the same room** (e.g. [💬 text-gene
The bot supports the following use-purposes:
- [💬 text-generation](#-text-generation): communicating with you via text (though certain models may "see" images as well)
- [💬 text-generation](#-text-generation): communicating with you via text (though certain models may also process images and files)
- [🦻 speech-to-text](#-speech-to-text): turning your voice messages into text
- [🗣️ text-to-speech](#%EF%B8%8F-text-to-speech): turning bot or users text messages into voice messages
- [🖌️ image-generation](#%EF%B8%8F-image-generation): generating images based on instructions
@@ -26,12 +26,14 @@ Text Generation is the bot's ability to **respond to users' messages with text**
![Screenshot of Text Generation - a user sends a message and the bot replies in a new conversation thread](./screenshots/text-generation.webp)
Some models also support vision, so you may be able to mix text and images in the same conversation.
Some models also support vision and document understanding, so you may be able to mix text, images, and files (PDFs, text documents, etc.) in the same conversation. Note that certain providers may not support all file types or may have issues with specific files (e.g. scanned/image-based PDFs). If a file is rejected by the provider, the conversation thread may become unusable — start a new thread to work around this.
In multi-user (group) rooms, to avoid disturbing the normal conversation between people, the bot is auto-configured to only respond to messages starting with the command prefix (`!bai`) or direct mentions via the [💬 Text Generation / 🗟 Prefix Requirement Type](./configuration/text-generation.md#-prefix-requirement-type) setting.
Normally, the bot only responds to allowed [👥 Users](./access.md#-users). In certain cases, it's useful for an allowed user to provoke the bot to respond even in foreign threads or reply chains. You can learn more about this feature in the [On-demand involvement](./features.md#on-demand-involvement) section below.
If needed, the bot can also attach sender metadata to conversation messages before sending them to the model, which can help the model distinguish between participants in multi-user rooms. See [🛠️ Configuration / 💬 Text Generation / 👤 Sender Context Mode](./configuration/text-generation.md#-sender-context-mode).
A few other features (like [🗣️ Text-to-Speech](#️-text-to-speech) and [🦻 Speech-to-Text](#-speech-to-text)) combine well with Text Generation, so you **don't necessarily need to communicate with the bot via text** (with [Seamless voice interaction](#seamless-voice-interaction), you can communicate only with voice).
You may also wish to see:

View File

@@ -1,7 +1,7 @@
base_url: https://api.openai.com/v1
api_key: YOUR_API_KEY_HERE
text_generation:
model_id: gpt-5.2
model_id: gpt-5.4
prompt: "You are a brief, but helpful bot called {{ baibot_name }} powered by the {{ baibot_model_id }} model. The date/time of this conversation's start is: {{ baibot_conversation_start_time_utc }}."
temperature: 1.0
# Reasoning models need to use `max_completion_tokens` instead of `max_response_tokens`.

View File

@@ -11,7 +11,7 @@ This is related to the [💬 Text Generation](./features.md#-text-generation) fe
If there's a text-generation handler agent configured, the bot **may** respond to messages sent in the room.
Some models also support vision, so you may be able to mix text and images in the same conversation.
Some models also support vision and document understanding, so you may be able to mix text, images, and files (PDFs, text documents, etc.) in the same conversation.
See screenshots of:

View File

@@ -11,7 +11,7 @@ user:
# Password-based login (traditional homeservers):
password: baibot
# Access token login (for MAS/OIDC-enabled homeservers):
# Access token login (for Matrix Authentication Service/OIDC-enabled homeservers):
# Generate a token via: mas-cli manage issue-compatibility-token <username> [device_id]
# access_token: null
# device_id: null
@@ -91,7 +91,7 @@ agents:
# base_url: https://api.openai.com/v1
# api_key: ""
# text_generation:
# model_id: gpt-5.2
# model_id: gpt-5.4
# prompt: "You are a brief, but helpful bot called {{ baibot_name }} powered by the {{ baibot_model_id }} model. The date/time of this conversation's start is: {{ baibot_conversation_start_time_utc }}."
# temperature: 1.0
# # Reasoning models need to use `max_completion_tokens` instead of `max_response_tokens`.

View File

@@ -1,6 +1,6 @@
services:
continuwuity:
image: forgejo.ellis.link/continuwuation/continuwuity:v0.5.6
image: forgejo.ellis.link/continuwuation/continuwuity:v0.5.9
user: "${UID}:${GID}"
restart: unless-stopped
cap_drop:

View File

@@ -1,6 +1,6 @@
services:
element-web:
image: ghcr.io/element-hq/element-web:v1.12.11
image: ghcr.io/element-hq/element-web:v1.12.17
user: "${UID}:${GID}"
restart: unless-stopped
environment:

View File

@@ -1,6 +1,6 @@
services:
ollama:
image: docker.io/ollama/ollama:0.17.6
image: docker.io/ollama/ollama:0.23.2
restart: unless-stopped
ports:
- "${SERVICE_OLLAMA_BIND_PORT_HTTP}:11434"

View File

@@ -14,7 +14,7 @@ services:
- /etc/passwd:/etc/passwd:ro
synapse:
image: ghcr.io/element-hq/synapse:v1.148.0
image: ghcr.io/element-hq/synapse:v1.152.1
user: "${UID}:${GID}"
restart: unless-stopped
entrypoint: python

View File

@@ -1,5 +1,5 @@
[tools]
prek = "0.3.2"
prek = "0.3.13"
[settings]
# Disable automatic trust prompts - we trust this config

4
rust-toolchain.toml Normal file
View File

@@ -0,0 +1,4 @@
[toolchain]
channel = "1.95.0"
components = ["rustfmt", "clippy"]
profile = "default"

View File

@@ -71,6 +71,7 @@ impl ControllerTrait for Controller {
let messages = vec![LLMMessage {
author: LLMAuthor::User,
sender_id: None,
content: LLMMessageContent::Text("Hello!".to_string()),
timestamp: chrono::Utc::now(),
}];
@@ -108,6 +109,7 @@ impl ControllerTrait for Controller {
} else {
Some(LLMMessage {
author: LLMAuthor::Prompt,
sender_id: None,
content: LLMMessageContent::Text(prompt_text),
timestamp: chrono::Utc::now(),
})

View File

@@ -28,6 +28,13 @@ pub(super) fn create_anthropic_message_request(llm_messages: Vec<LLMMessage>) ->
},
}]
}
LLMMessageContent::File(file_details) => {
tracing::warn!(
"The Anthropic provider's library does not support file/document content. This file message ({}) will be skipped.",
file_details.filename(),
);
continue;
}
};
let message = Message { role, content };

View File

@@ -84,7 +84,7 @@ impl Default for TextGenerationConfig {
}
fn default_text_model_id() -> String {
"gpt-5.2".to_owned()
"gpt-5.4".to_owned()
}
#[derive(Debug, Clone, Serialize, Deserialize, Default)]

View File

@@ -67,6 +67,7 @@ impl ControllerTrait for Controller {
let messages = vec![LLMMessage {
author: LLMAuthor::User,
sender_id: None,
content: LLMMessageContent::Text("Hello!".to_string()),
timestamp: chrono::Utc::now(),
}];
@@ -104,6 +105,7 @@ impl ControllerTrait for Controller {
} else {
Some(LLMMessage {
author: LLMAuthor::Prompt,
sender_id: None,
content: LLMMessageContent::Text(prompt_text),
timestamp: chrono::Utc::now(),
})

View File

@@ -1,6 +1,6 @@
use async_openai::types::responses::{
EasyInputContent, EasyInputMessage, ImageDetail, InputContent, InputImageContent, InputItem,
InputParam, MessageType, Role,
EasyInputContent, EasyInputMessage, ImageDetail, InputContent, InputFileArgs,
InputImageContent, InputItem, InputParam, MessageType, Role,
};
use crate::conversation::llm::{
@@ -35,12 +35,28 @@ pub fn convert_llm_messages_to_openai_response_input(
file_id: None,
})])
}
LLMMessageContent::File(file_details) => {
let file_data = format!(
"data:{};base64,{}",
file_details.mime,
base64_encode(&file_details.data)
);
let file_content = InputFileArgs::default()
.file_data(file_data)
.filename(file_details.filename())
.build()
.expect("Failed to build InputFileContent");
EasyInputContent::ContentList(vec![InputContent::InputFile(file_content)])
}
};
items.push(InputItem::EasyMessage(EasyInputMessage {
r#type: MessageType::Message,
role,
content,
phase: None,
}));
}

View File

@@ -64,6 +64,7 @@ impl ControllerTrait for Controller {
let messages = vec![LLMMessage {
author: LLMAuthor::User,
sender_id: None,
content: LLMMessageContent::Text("Hello!".to_string()),
timestamp: chrono::Utc::now(),
}];
@@ -101,6 +102,7 @@ impl ControllerTrait for Controller {
} else {
Some(LLMMessage {
author: LLMAuthor::Prompt,
sender_id: None,
content: LLMMessageContent::Text(prompt_text),
timestamp: chrono::Utc::now(),
})

View File

@@ -40,6 +40,12 @@ fn convert_llm_message_to_openai_message(llm_message: LLMMessage) -> Option<Mess
);
None
}
LLMMessageContent::File(_file_details) => {
tracing::warn!(
"The OpenAI-compat provider's library does not support file content. This file message will be skipped."
);
None
}
}
}

View File

@@ -3,7 +3,8 @@ use crate::{
entity::roomconfig::{
SpeechToTextFlowType, SpeechToTextMessageTypeForNonThreadedOnlyTranscribedMessages,
TextGenerationAutoUsage, TextGenerationPrefixRequirementType,
TextToSpeechBotMessagesFlowType, TextToSpeechUserMessagesFlowType,
TextGenerationSenderContextMode, TextToSpeechBotMessagesFlowType,
TextToSpeechUserMessagesFlowType,
},
};
@@ -48,6 +49,9 @@ pub enum ConfigTextGenerationSettingRelatedControllerType {
GetTemperatureOverride,
SetTemperatureOverride(Option<f32>),
GetSenderContextMode,
SetSenderContextMode(Option<TextGenerationSenderContextMode>),
}
#[derive(Debug, PartialEq)]

View File

@@ -163,6 +163,26 @@ fn determine_controller() {
),
)),
},
TestCase {
name: "per-room text-generation/sender-context-mode getter",
input: "room text-generation sender-context-mode",
expected: super::ControllerType::Config(controller_type::ConfigControllerType::SettingsRelated(
controller_type::SettingsStorageSource::Room,
controller_type::ConfigSettingRelatedControllerType::TextGeneration(
controller_type::ConfigTextGenerationSettingRelatedControllerType::GetSenderContextMode,
),
)),
},
TestCase {
name: "global text-generation/sender-context-mode getter",
input: "global text-generation sender-context-mode",
expected: super::ControllerType::Config(controller_type::ConfigControllerType::SettingsRelated(
controller_type::SettingsStorageSource::Global,
controller_type::ConfigSettingRelatedControllerType::TextGeneration(
controller_type::ConfigTextGenerationSettingRelatedControllerType::GetSenderContextMode,
),
)),
},
TestCase {
name: "per-room text-to-speech/speed-override getter",
input: "room text-to-speech speed-override",

View File

@@ -3,7 +3,10 @@ mod tests;
use crate::{
controller::ControllerType,
entity::roomconfig::{TextGenerationAutoUsage, TextGenerationPrefixRequirementType},
entity::roomconfig::{
TextGenerationAutoUsage, TextGenerationPrefixRequirementType,
TextGenerationSenderContextMode,
},
strings,
};
@@ -197,5 +200,43 @@ pub(super) fn determine(
);
}
if let Some(remaining_text) = text.strip_prefix("sender-context-mode") {
let remaining_text = remaining_text.trim();
if !remaining_text.is_empty() {
return Err(ControllerType::Error(
strings::cfg::configuration_getter_used_with_extra_text(
"sender-context-mode",
remaining_text,
)
.to_owned(),
));
}
return Ok(ConfigTextGenerationSettingRelatedControllerType::GetSenderContextMode);
}
if let Some(value_string) = text.strip_prefix("set-sender-context-mode") {
let value_string = value_string.trim().to_owned();
let value_choice = if value_string.is_empty() {
None
} else {
let value_choice =
TextGenerationSenderContextMode::from_str(&value_string.to_lowercase());
if value_choice.is_none() {
return Err(ControllerType::Error(
strings::cfg::configuration_value_unrecognized(&value_string).to_owned(),
));
}
value_choice
};
return Ok(
ConfigTextGenerationSettingRelatedControllerType::SetSenderContextMode(value_choice),
);
}
Err(ControllerType::Unknown)
}

View File

@@ -90,6 +90,74 @@ fn determine_controller_context_management() {
}
}
#[test]
fn determine_controller_sender_context() {
use super::ConfigTextGenerationSettingRelatedControllerType;
use super::ControllerType;
use crate::entity::roomconfig::TextGenerationSenderContextMode;
struct TestCase {
name: &'static str,
input: &'static str,
expected: Result<ConfigTextGenerationSettingRelatedControllerType, ControllerType>,
}
let test_cases = vec![
TestCase {
name: "sender-context-mode getter ok",
input: "sender-context-mode",
expected: Ok(ConfigTextGenerationSettingRelatedControllerType::GetSenderContextMode),
},
TestCase {
name: "sender-context-mode getter extra args",
input: "sender-context-mode some values here",
expected: Err(ControllerType::Error(
crate::strings::cfg::configuration_getter_used_with_extra_text(
"sender-context-mode",
"some values here",
),
)),
},
TestCase {
name: "sender-context-mode setter matrix_user_id",
input: "set-sender-context-mode matrix_user_id",
expected: Ok(
ConfigTextGenerationSettingRelatedControllerType::SetSenderContextMode(Some(
TextGenerationSenderContextMode::MatrixUserId,
)),
),
},
TestCase {
name: "sender-context-mode setter uppercase",
input: "set-sender-context-mode MATRIX_USER_ID_AND_TIMESTAMP",
expected: Ok(
ConfigTextGenerationSettingRelatedControllerType::SetSenderContextMode(Some(
TextGenerationSenderContextMode::MatrixUserIdAndTimestamp,
)),
),
},
TestCase {
name: "sender-context-mode setter invalid",
input: "set-sender-context-mode non-Enum-Value",
expected: Err(ControllerType::Error(
crate::strings::cfg::configuration_value_unrecognized("non-Enum-Value"),
)),
},
TestCase {
name: "sender-context-mode unsetter",
input: "set-sender-context-mode",
expected: Ok(
ConfigTextGenerationSettingRelatedControllerType::SetSenderContextMode(None),
),
},
];
for test_case in test_cases {
let result = super::determine(test_case.input);
assert_eq!(result, test_case.expected, "Test case: {}", test_case.name);
}
}
#[test]
fn determine_controller_prefix_requirement_type() {
use super::ConfigTextGenerationSettingRelatedControllerType;

View File

@@ -1,5 +1,6 @@
use crate::entity::roomconfig::{
RoomSettings, TextGenerationAutoUsage, TextGenerationPrefixRequirementType,
TextGenerationSenderContextMode,
};
use crate::{Bot, entity::MessageContext};
@@ -151,5 +152,38 @@ pub(super) async fn dispatch(
}
}
}
ConfigTextGenerationSettingRelatedControllerType::GetSenderContextMode => {
let value = &room_settings.text_generation.sender_context_mode;
setting_get::<TextGenerationSenderContextMode>(bot, message_context, value).await
}
ConfigTextGenerationSettingRelatedControllerType::SetSenderContextMode(value) => {
let value = value.to_owned();
let setter_callback = Box::new(move |room_settings: &mut RoomSettings| {
room_settings.text_generation.sender_context_mode = value;
});
match config_type {
SettingsStorageSource::Room => {
room_setting_set::<TextGenerationSenderContextMode>(
bot,
message_context,
&value,
setter_callback,
)
.await
}
SettingsStorageSource::Global => {
global_setting_set::<TextGenerationSenderContextMode>(
bot,
message_context,
&value,
setter_callback,
)
.await
}
}
}
}
}

View File

@@ -7,7 +7,8 @@ use crate::{
roomconfig::{
SpeechToTextFlowType, SpeechToTextMessageTypeForNonThreadedOnlyTranscribedMessages,
TextGenerationAutoUsage, TextGenerationPrefixRequirementType,
TextToSpeechBotMessagesFlowType, TextToSpeechUserMessagesFlowType,
TextGenerationSenderContextMode, TextToSpeechBotMessagesFlowType,
TextToSpeechUserMessagesFlowType,
},
},
strings,
@@ -233,6 +234,46 @@ fn build_section_text_generation(command_prefix: &str, bot_username: &str) -> St
));
message.push_str("\n\n");
// Sender Context
message.push_str(&format!(
"#### {}",
strings::help::cfg::text_generation_sender_context_heading()
));
message.push_str("\n\n");
message.push_str(&strings::help::cfg::text_generation_sender_context_intro());
message.push('\n');
message.push_str(
&strings::help::cfg::the_following_configuration_values_are_recognized(
TextGenerationSenderContextMode::choices(),
),
);
message.push_str("\n\n");
message.push_str(&format!(
"- {}",
&strings::help::cfg::current_setting_show(
command_prefix,
"text-generation sender-context-mode"
)
));
message.push('\n');
message.push_str(&format!(
"- {}",
&strings::help::cfg::current_setting_set(
command_prefix,
"text-generation set-sender-context-mode VALUE"
)
));
message.push('\n');
message.push_str(&format!(
"- {}",
&strings::help::cfg::current_setting_unset(
command_prefix,
"text-generation set-sender-context-mode"
)
));
message.push_str("\n\n");
// Prompt override
message.push_str(&format!(

View File

@@ -359,6 +359,33 @@ async fn generate_text_generation_section(
),
);
// Sender Context
let effective_sender_context = room_config_context.text_generation_sender_context_mode();
let room_config_sender_context = room_config_context
.room_config
.settings
.text_generation
.sender_context_mode;
let global_config_sender_context = room_config_context
.global_config
.fallback_room_settings
.text_generation
.sender_context_mode;
let sender_context_set_where = if room_config_sender_context.is_some() {
strings::cfg::status_badge_set_in_room_config()
} else if global_config_sender_context.is_some() {
strings::cfg::status_badge_set_in_global_config()
} else {
strings::cfg::status_badge_using_hardcoded_default()
};
message.push_str(&strings::cfg::status_text_generation_entry_sender_context(
effective_sender_context,
sender_context_set_where,
));
// Prompt override
let text_agent_prompt = if let Some(text_generation_agent) = &text_generation_agent {

View File

@@ -15,7 +15,8 @@ use crate::conversation::matrix::MatrixMessageProcessingParams;
use crate::entity::MessagePayload;
use crate::entity::roomconfig::{
SpeechToTextFlowType, SpeechToTextMessageTypeForNonThreadedOnlyTranscribedMessages,
TextToSpeechBotMessagesFlowType, TextToSpeechUserMessagesFlowType,
TextGenerationSenderContextMode, TextToSpeechBotMessagesFlowType,
TextToSpeechUserMessagesFlowType,
};
use crate::strings;
use crate::utils::text_to_speech::create_transcribed_message_text;
@@ -23,6 +24,7 @@ use crate::{
Bot,
conversation::{
create_llm_conversation_for_matrix_reply_chain, create_llm_conversation_for_matrix_thread,
llm::{Author, Conversation, MessageContent},
matrix::create_list_of_bot_user_prefixes_to_strip,
},
entity::MessageContext,
@@ -41,6 +43,8 @@ pub enum ChatCompletionControllerType {
Image,
File,
ThreadMention,
ReplyMention,
}
@@ -419,7 +423,8 @@ async fn handle_stage_text_generation(
| ChatCompletionControllerType::TextMention
| ChatCompletionControllerType::TextDirect
| ChatCompletionControllerType::Audio
| ChatCompletionControllerType::Image => {
| ChatCompletionControllerType::Image
| ChatCompletionControllerType::File => {
Some(message_context.combined_admin_and_user_regexes())
}
@@ -483,6 +488,13 @@ async fn handle_stage_text_generation(
}
};
let conversation = inject_sender_context(
conversation,
message_context
.room_config_context()
.text_generation_sender_context_mode(),
);
tracing::debug!(
agent_id = agent.identifier().as_string(),
provider = format!("{}", agent.definition().provider.clone()),
@@ -758,3 +770,238 @@ async fn generate_and_send_tts_for_message(
)
.await
}
fn inject_sender_context(
conversation: Conversation,
sender_context_mode: TextGenerationSenderContextMode,
) -> Conversation {
if sender_context_mode == TextGenerationSenderContextMode::Disabled {
return conversation;
}
let include_timestamp =
sender_context_mode == TextGenerationSenderContextMode::MatrixUserIdAndTimestamp;
let messages = conversation
.messages
.into_iter()
.map(|mut message| {
if message.author == Author::Prompt {
return message;
}
let Some(sender_id) = &message.sender_id else {
return message;
};
if let MessageContent::Text(ref mut text) = message.content {
*text = if include_timestamp {
let timestamp = message.timestamp.format("%Y-%m-%dT%H:%M:%SZ");
format!("[sender={} sent_at={}] {}", sender_id, timestamp, text)
} else {
format!("[sender={}] {}", sender_id, text)
};
}
message
})
.collect();
Conversation { messages }
}
#[cfg(test)]
mod sender_context_tests {
use super::inject_sender_context;
use crate::conversation::llm::{Author, Conversation, ImageDetails, Message, MessageContent};
use crate::entity::roomconfig::TextGenerationSenderContextMode;
use chrono::{TimeZone, Utc};
use mxlink::matrix_sdk::ruma::events::room::message::ImageMessageEventContent;
use mxlink::matrix_sdk::ruma::{OwnedMxcUri, OwnedUserId};
use mxlink::mime;
#[test]
fn test_inject_sender_context_prefixes_text_messages() {
let timestamp = Utc.with_ymd_and_hms(2026, 3, 23, 14, 30, 0).unwrap();
let user_id = OwnedUserId::try_from("@alice:example.com").unwrap();
let conversation = Conversation {
messages: vec![Message {
author: Author::User,
sender_id: Some(user_id),
timestamp,
content: MessageContent::Text("Hello bot".to_string()),
}],
};
let result = inject_sender_context(
conversation,
TextGenerationSenderContextMode::MatrixUserIdAndTimestamp,
);
assert_eq!(result.messages.len(), 1);
assert_eq!(
result.messages[0].content,
MessageContent::Text(
"[sender=@alice:example.com sent_at=2026-03-23T14:30:00Z] Hello bot".to_string()
)
);
}
#[test]
fn test_inject_sender_context_can_prefix_without_timestamp() {
let timestamp = Utc.with_ymd_and_hms(2026, 3, 23, 14, 30, 0).unwrap();
let user_id = OwnedUserId::try_from("@alice:example.com").unwrap();
let conversation = Conversation {
messages: vec![Message {
author: Author::User,
sender_id: Some(user_id),
timestamp,
content: MessageContent::Text("Hello bot".to_string()),
}],
};
let result =
inject_sender_context(conversation, TextGenerationSenderContextMode::MatrixUserId);
assert_eq!(result.messages.len(), 1);
assert_eq!(
result.messages[0].content,
MessageContent::Text("[sender=@alice:example.com] Hello bot".to_string())
);
}
#[test]
fn test_inject_sender_context_prefixes_assistant_messages() {
let timestamp = Utc.with_ymd_and_hms(2026, 3, 23, 14, 30, 0).unwrap();
let user_id = OwnedUserId::try_from("@baibot:example.com").unwrap();
let conversation = Conversation {
messages: vec![Message {
author: Author::Assistant,
sender_id: Some(user_id),
timestamp,
content: MessageContent::Text("Hello human".to_string()),
}],
};
let result = inject_sender_context(
conversation,
TextGenerationSenderContextMode::MatrixUserIdAndTimestamp,
);
assert_eq!(result.messages.len(), 1);
assert_eq!(
result.messages[0].content,
MessageContent::Text(
"[sender=@baibot:example.com sent_at=2026-03-23T14:30:00Z] Hello human".to_string()
)
);
}
#[test]
fn test_inject_sender_context_skips_prompt_messages() {
let timestamp = Utc.with_ymd_and_hms(2026, 3, 23, 14, 30, 0).unwrap();
let conversation = Conversation {
messages: vec![Message {
author: Author::Prompt,
sender_id: None,
timestamp,
content: MessageContent::Text("You are a bot".to_string()),
}],
};
let result =
inject_sender_context(conversation, TextGenerationSenderContextMode::MatrixUserId);
assert_eq!(
result.messages[0].content,
MessageContent::Text("You are a bot".to_string())
);
}
#[test]
fn test_inject_sender_context_skips_messages_without_sender_id() {
let timestamp = Utc.with_ymd_and_hms(2026, 3, 23, 14, 30, 0).unwrap();
let conversation = Conversation {
messages: vec![Message {
author: Author::User,
sender_id: None,
timestamp,
content: MessageContent::Text("Transcribed text".to_string()),
}],
};
let result = inject_sender_context(
conversation,
TextGenerationSenderContextMode::MatrixUserIdAndTimestamp,
);
assert_eq!(
result.messages[0].content,
MessageContent::Text("Transcribed text".to_string())
);
}
#[test]
fn test_inject_sender_context_leaves_non_text_content_unchanged() {
let timestamp = Utc.with_ymd_and_hms(2026, 3, 23, 14, 30, 0).unwrap();
let user_id = OwnedUserId::try_from("@alice:example.com").unwrap();
let image_event_content = ImageMessageEventContent::plain(
"image.png".to_string(),
OwnedMxcUri::from("mxc://example.com/1234567890"),
);
let conversation = Conversation {
messages: vec![Message {
author: Author::User,
sender_id: Some(user_id),
timestamp,
content: MessageContent::Image(ImageDetails::new(
image_event_content.clone(),
mime::IMAGE_PNG,
vec![],
)),
}],
};
let result = inject_sender_context(
conversation,
TextGenerationSenderContextMode::MatrixUserIdAndTimestamp,
);
assert_eq!(
result.messages[0].content,
MessageContent::Image(ImageDetails::new(
image_event_content,
mime::IMAGE_PNG,
vec![]
))
);
}
#[test]
fn test_inject_sender_context_none_leaves_text_unchanged() {
let timestamp = Utc.with_ymd_and_hms(2026, 3, 23, 14, 30, 0).unwrap();
let user_id = OwnedUserId::try_from("@alice:example.com").unwrap();
let conversation = Conversation {
messages: vec![Message {
author: Author::User,
sender_id: Some(user_id),
timestamp,
content: MessageContent::Text("Hello bot".to_string()),
}],
};
let result = inject_sender_context(conversation, TextGenerationSenderContextMode::Disabled);
assert_eq!(
result.messages[0].content,
MessageContent::Text("Hello bot".to_string())
);
}
}

View File

@@ -58,6 +58,18 @@ pub fn determine_controller(
)
}
}
MessagePayload::File(_file_message_content) => {
let prefix_requirement_type = message_context
.room_config_context()
.text_generation_prefix_requirement_type();
match prefix_requirement_type {
TextGenerationPrefixRequirementType::CommandPrefix => ControllerType::Ignore,
TextGenerationPrefixRequirementType::No => {
ControllerType::ChatCompletion(ChatCompletionControllerType::File)
}
}
}
MessagePayload::Audio(_) => {
ControllerType::ChatCompletion(ChatCompletionControllerType::Audio)
}

View File

@@ -64,6 +64,7 @@ mod tests {
original_prompt: "Generate a picture of a dog",
messages: vec![Message {
author: Author::User,
sender_id: None,
content: MessageContent::Text("Must be blue".to_owned()),
timestamp,
}],
@@ -75,16 +76,19 @@ mod tests {
messages: vec![
Message {
author: Author::User,
sender_id: None,
content: MessageContent::Text("Must be blue".to_owned()),
timestamp,
},
Message {
author: Author::Assistant,
sender_id: None,
content: MessageContent::Text("Whatever".to_owned()),
timestamp,
},
Message {
author: Author::User,
sender_id: None,
content: MessageContent::Text(
"Must be 3-legged.\nMust be flying.".to_owned(),
),
@@ -99,21 +103,25 @@ mod tests {
messages: vec![
Message {
author: Author::User,
sender_id: None,
content: MessageContent::Text("Must be blue".to_owned()),
timestamp,
},
Message {
author: Author::Assistant,
sender_id: None,
content: MessageContent::Text("Whatever".to_owned()),
timestamp,
},
Message {
author: Author::User,
sender_id: None,
content: MessageContent::Text("Again".to_owned()),
timestamp,
},
Message {
author: Author::User,
sender_id: None,
content: MessageContent::Text("again".to_owned()),
timestamp,
},

View File

@@ -1,5 +1,8 @@
use chrono::{DateTime, Utc};
use mxlink::matrix_sdk::ruma::events::room::message::ImageMessageEventContent;
use mxlink::matrix_sdk::ruma::OwnedUserId;
use mxlink::matrix_sdk::ruma::events::room::message::{
FileMessageEventContent, ImageMessageEventContent,
};
use mxlink::mime::Mime;
use crate::agent::provider::ImageSource;
@@ -14,6 +17,7 @@ pub enum Author {
#[derive(Debug, Clone)]
pub struct Message {
pub author: Author,
pub sender_id: Option<OwnedUserId>,
pub timestamp: DateTime<Utc>,
pub content: MessageContent,
}
@@ -48,10 +52,35 @@ impl From<ImageDetails> for ImageSource {
}
}
#[derive(Debug, Clone)]
pub struct FileDetails {
pub event_content: FileMessageEventContent,
pub mime: Mime,
pub data: Vec<u8>,
}
impl FileDetails {
pub fn new(event_content: FileMessageEventContent, mime: Mime, data: Vec<u8>) -> Self {
Self {
event_content,
mime,
data,
}
}
pub fn filename(&self) -> String {
self.event_content
.filename
.clone()
.unwrap_or(self.event_content.body.clone())
}
}
#[derive(Debug, Clone)]
pub enum MessageContent {
Text(String),
Image(ImageDetails),
File(FileDetails),
}
impl PartialEq for MessageContent {
@@ -62,6 +91,7 @@ impl PartialEq for MessageContent {
// We can probably do better than this by inspecting `.event_conten1t.source`, but for now this is good enough.
a.filename() == b.filename()
}
(MessageContent::File(a), MessageContent::File(b)) => a.filename() == b.filename(),
_ => false,
}
}
@@ -76,6 +106,11 @@ impl Conversation {
///
/// Certain models (like Anthropic) cannot tolerate consecutive messages by the same author,
/// so combining them helps avoid issues.
///
/// When multiple text messages by the same author are merged, the resulting message keeps a
/// `sender_id` only if all merged messages came from the same sender. Mixed-sender merges are
/// possible for user turns in multi-user rooms, so `sender_id` is cleared in that case to
/// avoid incorrectly attributing the whole merged turn to the first sender.
/// See: https://github.com/etkecc/baibot/issues/13
pub fn combine_consecutive_messages(&self) -> Conversation {
// We'll likely get fewer messages, but let's reserve the maximum we expect.
@@ -106,6 +141,10 @@ impl Conversation {
text.push('\n');
text.push_str(message_text_content);
}
if last_message.sender_id != message.sender_id {
last_message.sender_id = None;
}
}
Conversation {
@@ -122,7 +161,7 @@ impl Conversation {
mod tests {
use super::*;
use chrono::{TimeZone, Utc};
use mxlink::matrix_sdk::ruma::OwnedMxcUri;
use mxlink::matrix_sdk::ruma::{OwnedMxcUri, OwnedUserId};
use mxlink::mime;
#[test]
@@ -145,21 +184,25 @@ mod tests {
// User's turn
Message {
author: Author::User,
sender_id: None,
content: MessageContent::Text("Hello".to_string()),
timestamp: timestamp_1,
},
Message {
author: Author::User,
sender_id: None,
content: MessageContent::Text("How are you?".to_string()),
timestamp: timestamp_2,
},
Message {
author: Author::User,
sender_id: None,
content: MessageContent::Text("I'm OK, btw.".to_string()),
timestamp: timestamp_3,
},
Message {
author: Author::User,
sender_id: None,
content: MessageContent::Image(ImageDetails::new(
image_event_content.clone(),
mime::IMAGE_PNG,
@@ -169,28 +212,33 @@ mod tests {
},
Message {
author: Author::User,
sender_id: None,
content: MessageContent::Text("Above is an image.".to_string()),
timestamp: timestamp_4,
},
Message {
author: Author::User,
sender_id: None,
content: MessageContent::Text("Would you take a look at it?".to_string()),
timestamp: timestamp_4,
},
// Assistant's turn
Message {
author: Author::Assistant,
sender_id: None,
content: MessageContent::Text("Hi there!".to_string()),
timestamp: timestamp_2,
},
Message {
author: Author::Assistant,
sender_id: None,
content: MessageContent::Text("I'm doing well, thank you.".to_string()),
timestamp: timestamp_3,
},
// User's turn
Message {
author: Author::User,
sender_id: None,
content: MessageContent::Text("That's great!".to_string()),
timestamp: timestamp_3,
},
@@ -239,4 +287,39 @@ mod tests {
);
assert_eq!(conversation.messages[4].timestamp, timestamp_3);
}
#[test]
fn combine_consecutive_messages_clears_sender_id_for_mixed_sender_turns() {
let timestamp_1 = Utc.with_ymd_and_hms(2024, 9, 20, 18, 34, 15).unwrap();
let timestamp_2 = Utc.with_ymd_and_hms(2024, 9, 20, 18, 34, 16).unwrap();
let sender_1 = OwnedUserId::try_from("@alice:example.com").unwrap();
let sender_2 = OwnedUserId::try_from("@bob:example.com").unwrap();
let conversation = Conversation {
messages: vec![
Message {
author: Author::User,
sender_id: Some(sender_1),
content: MessageContent::Text("Hello".to_string()),
timestamp: timestamp_1,
},
Message {
author: Author::User,
sender_id: Some(sender_2),
content: MessageContent::Text("Hi there".to_string()),
timestamp: timestamp_2,
},
],
};
let conversation = conversation.combine_consecutive_messages();
assert_eq!(conversation.messages.len(), 1);
assert_eq!(conversation.messages[0].sender_id, None);
assert_eq!(
conversation.messages[0].content,
MessageContent::Text("Hello\nHi there".to_string())
);
assert_eq!(conversation.messages[0].timestamp, timestamp_1);
}
}

View File

@@ -20,6 +20,7 @@ fn test_messages_by_the_bot_are_identified_correctly() {
let llm_message = convert_matrix_message_to_llm_message(&matrix_message, &bot_user_id).unwrap();
assert_eq!(llm_message.author, Author::Assistant);
assert_eq!(llm_message.sender_id, Some(bot_user_id.clone()));
assert_eq!(
llm_message.content,
MessageContent::Text("Hello!".to_string())
@@ -47,6 +48,7 @@ fn test_notice_messages_by_bot_with_speech_to_text_prefix_are_cleaned_up_and_con
let llm_message = convert_matrix_message_to_llm_message(&matrix_message, &bot_user_id).unwrap();
assert_eq!(llm_message.author, Author::User);
assert_eq!(llm_message.sender_id, None);
assert_eq!(
llm_message.content,
MessageContent::Text(source_message_text.to_string())
@@ -75,6 +77,30 @@ fn test_notice_error_messages_by_bot_are_ignored() {
assert!(llm_message.is_none());
}
#[test]
fn test_user_messages_preserve_sender_id() {
let bot_user_id =
OwnedUserId::try_from("@bot:example.com").expect("Failed to parse bot user ID");
let user_id = OwnedUserId::try_from("@alice:example.com").expect("Failed to parse user ID");
let matrix_message = super::super::matrix::MatrixMessage {
sender_id: user_id.clone(),
content: super::super::matrix::MatrixMessageContent::Text("Hello!".to_owned()),
mentioned_users: vec![],
timestamp: chrono::Utc::now(),
};
let llm_message = convert_matrix_message_to_llm_message(&matrix_message, &bot_user_id).unwrap();
assert_eq!(llm_message.author, Author::User);
assert_eq!(llm_message.sender_id, Some(user_id));
assert_eq!(
llm_message.content,
MessageContent::Text("Hello!".to_string())
);
}
#[test]
fn test_other_notice_messages_by_the_bot_are_ignored() {
// Also see `test_notice_error_messages_by_bot_are_ignored()`.

View File

@@ -1,15 +1,15 @@
use tiktoken_rs::CoreBPE;
use tiktoken_rs::get_bpe_from_tokenizer;
use tiktoken_rs::bpe_for_tokenizer;
use tiktoken_rs::tokenizer;
use super::{Author, Message, MessageContent};
fn get_bpe_for_model(model: &str) -> CoreBPE {
fn get_bpe_for_model(model: &str) -> &'static CoreBPE {
let tokenizer = tokenizer::get_tokenizer(model)
.or_else(|| tokenizer::get_tokenizer("gpt-4"))
.unwrap();
get_bpe_from_tokenizer(tokenizer).unwrap()
bpe_for_tokenizer(tokenizer).unwrap()
}
pub fn shorten_messages_list_to_context_size(
@@ -26,7 +26,7 @@ pub fn shorten_messages_list_to_context_size(
// We want to retain the prompt in all cases, so we always count it first.
// We also always reserve enough tokens for the maximum response we expect.
let mut current_context_length: u32 = if let Some(prompt_message) = prompt_message {
calculate_token_size_for_message(&bpe, model, prompt_message)
calculate_token_size_for_message(bpe, model, prompt_message)
+ max_response_tokens.unwrap_or(0)
} else {
0
@@ -37,7 +37,7 @@ pub fn shorten_messages_list_to_context_size(
let mut messages_to_keep: Vec<Message> = Vec::new();
for message in messages {
let tokens_for_message = calculate_token_size_for_message(&bpe, model, &message);
let tokens_for_message = calculate_token_size_for_message(bpe, model, &message);
if current_context_length + tokens_for_message > max_context_tokens {
break;
@@ -74,6 +74,7 @@ fn calculate_token_size_for_message(bpe: &CoreBPE, model: &str, message: &Messag
let text_length = match &message.content {
MessageContent::Text(text) => bpe.encode_with_special_tokens(text).len() as i32,
MessageContent::Image(..) => 0,
MessageContent::File(..) => 0,
};
(text_length + role_length + tokens_per_message + tokens_per_name) as u32
@@ -88,11 +89,12 @@ pub mod test {
let message = super::Message {
author: super::Author::User,
sender_id: None,
content: super::MessageContent::Text("Hello there!".to_string()),
timestamp: chrono::Utc::now(),
};
let tokens = super::calculate_token_size_for_message(&bpe, model, &message);
let tokens = super::calculate_token_size_for_message(bpe, model, &message);
assert_eq!(8, tokens);
}
@@ -107,6 +109,7 @@ pub mod test {
let prompt = super::Message {
author: super::Author::Prompt,
sender_id: None,
content: super::MessageContent::Text("You are a bot!".to_string()),
timestamp: chrono::Utc::now(),
};
@@ -114,13 +117,14 @@ pub mod test {
assert_eq!(
prompt_length,
super::calculate_token_size_for_message(&bpe, model, &prompt)
super::calculate_token_size_for_message(bpe, model, &prompt)
);
let mut conversation_messages = Vec::new();
let first = super::Message {
author: super::Author::User,
sender_id: None,
content: super::MessageContent::Text("Hello there!".to_string()),
timestamp: chrono::Utc::now(),
};
@@ -128,13 +132,14 @@ pub mod test {
assert_eq!(
first_length,
super::calculate_token_size_for_message(&bpe, model, &first)
super::calculate_token_size_for_message(bpe, model, &first)
);
conversation_messages.push(first);
let second = super::Message {
author: super::Author::Assistant,
sender_id: None,
content: super::MessageContent::Text("Hello!".to_string()),
timestamp: chrono::Utc::now(),
};
@@ -142,13 +147,14 @@ pub mod test {
assert_eq!(
second_length,
super::calculate_token_size_for_message(&bpe, model, &second)
super::calculate_token_size_for_message(bpe, model, &second)
);
conversation_messages.push(second);
let third = super::Message {
author: super::Author::User,
sender_id: None,
content: super::MessageContent::Text(
"This is the 3rd message in this conversation. It shall be preserved.".to_owned(),
),
@@ -158,13 +164,14 @@ pub mod test {
assert_eq!(
third_length,
super::calculate_token_size_for_message(&bpe, model, &third)
super::calculate_token_size_for_message(bpe, model, &third)
);
conversation_messages.push(third.clone());
let forth = super::Message {
author: super::Author::Assistant,
sender_id: None,
content: super::MessageContent::Text(
"This is yet another message that shall be preserved.".to_owned(),
),
@@ -174,7 +181,7 @@ pub mod test {
assert_eq!(
forth_length,
super::calculate_token_size_for_message(&bpe, model, &forth)
super::calculate_token_size_for_message(bpe, model, &forth)
);
conversation_messages.push(forth.clone());
@@ -212,6 +219,7 @@ pub mod test {
let prompt = super::Message {
author: super::Author::User,
sender_id: None,
content: super::MessageContent::Text("あなたはボットです。".to_string()),
timestamp: chrono::Utc::now(),
};
@@ -219,13 +227,14 @@ pub mod test {
assert_eq!(
prompt_length,
super::calculate_token_size_for_message(&bpe, model, &prompt)
super::calculate_token_size_for_message(bpe, model, &prompt)
);
let mut conversation_messages = Vec::new();
let first = super::Message {
author: super::Author::User,
sender_id: None,
content: super::MessageContent::Text("こんにちは!".to_string()),
timestamp: chrono::Utc::now(),
};
@@ -233,13 +242,14 @@ pub mod test {
assert_eq!(
first_length,
super::calculate_token_size_for_message(&bpe, model, &first)
super::calculate_token_size_for_message(bpe, model, &first)
);
conversation_messages.push(first);
let second = super::Message {
author: super::Author::Assistant,
sender_id: None,
content: super::MessageContent::Text("こんにちは。今日は元気ですか。".to_string()),
timestamp: chrono::Utc::now(),
};
@@ -247,13 +257,14 @@ pub mod test {
assert_eq!(
second_length,
super::calculate_token_size_for_message(&bpe, model, &second)
super::calculate_token_size_for_message(bpe, model, &second)
);
conversation_messages.push(second);
let third = super::Message {
author: super::Author::User,
sender_id: None,
content: super::MessageContent::Text(
"これは第3のメッセージなので、保存されます。".to_string(),
),
@@ -263,13 +274,14 @@ pub mod test {
assert_eq!(
third_length,
super::calculate_token_size_for_message(&bpe, model, &third)
super::calculate_token_size_for_message(bpe, model, &third)
);
conversation_messages.push(third.clone());
let forth = super::Message {
author: super::Author::Assistant,
sender_id: None,
content: super::MessageContent::Text(
"これはもう一つの保存されますメッセージです。".to_string(),
),
@@ -279,7 +291,7 @@ pub mod test {
assert_eq!(
forth_length,
super::calculate_token_size_for_message(&bpe, model, &forth)
super::calculate_token_size_for_message(bpe, model, &forth)
);
conversation_messages.push(forth.clone());

View File

@@ -1,6 +1,6 @@
use mxlink::matrix_sdk::ruma::OwnedUserId;
use super::entity::{Author, ImageDetails, Message, MessageContent};
use super::entity::{Author, FileDetails, ImageDetails, Message, MessageContent};
use crate::conversation::matrix::{MatrixMessage, MatrixMessageContent};
use crate::utils::text_to_speech as text_to_speech_utils;
@@ -17,14 +17,17 @@ pub fn convert_matrix_message_to_llm_message(
fn convert_bot_message(matrix_message: &MatrixMessage) -> Option<Message> {
match &matrix_message.content {
MatrixMessageContent::Text(text) => {
convert_bot_text_message(text, &matrix_message.timestamp)
}
MatrixMessageContent::Text(text) => convert_bot_text_message(
text,
&matrix_message.timestamp,
matrix_message.sender_id.clone(),
),
MatrixMessageContent::Notice(text) => {
convert_bot_notice_message(text, &matrix_message.timestamp)
}
MatrixMessageContent::Image(image_content, mime_type, media_bytes) => Some(Message {
author: Author::Assistant,
sender_id: Some(matrix_message.sender_id.clone()),
content: MessageContent::Image(ImageDetails::new(
image_content.clone(),
mime_type.clone(),
@@ -32,15 +35,27 @@ fn convert_bot_message(matrix_message: &MatrixMessage) -> Option<Message> {
)),
timestamp: matrix_message.timestamp.to_owned(),
}),
MatrixMessageContent::File(file_content, mime_type, media_bytes) => Some(Message {
author: Author::Assistant,
sender_id: Some(matrix_message.sender_id.clone()),
content: MessageContent::File(FileDetails::new(
file_content.clone(),
mime_type.clone(),
media_bytes.clone(),
)),
timestamp: matrix_message.timestamp.to_owned(),
}),
}
}
fn convert_bot_text_message(
text: &str,
timestamp: &chrono::DateTime<chrono::Utc>,
sender_id: OwnedUserId,
) -> Option<Message> {
Some(Message {
author: Author::Assistant,
sender_id: Some(sender_id),
content: MessageContent::Text(text.to_owned()),
timestamp: timestamp.to_owned(),
})
@@ -59,8 +74,10 @@ fn convert_bot_notice_message(
if let Some(text) = text_to_speech_utils::parse_transcribed_message_text(text) {
// This is a transcription message. We remove the prefix and consider it as a message sent by the user.
// sender_id is None because the original speaker is unknown.
return Some(Message {
author: Author::User,
sender_id: None,
content: MessageContent::Text(text.to_owned()),
timestamp: timestamp.to_owned(),
});
@@ -73,16 +90,19 @@ fn convert_user_message(matrix_message: &MatrixMessage) -> Option<Message> {
match &matrix_message.content {
MatrixMessageContent::Text(text) => Some(Message {
author: Author::User,
sender_id: Some(matrix_message.sender_id.clone()),
content: MessageContent::Text(text.clone()),
timestamp: matrix_message.timestamp.to_owned(),
}),
MatrixMessageContent::Notice(text) => Some(Message {
author: Author::User,
sender_id: Some(matrix_message.sender_id.clone()),
content: MessageContent::Text(text.clone()),
timestamp: matrix_message.timestamp.to_owned(),
}),
MatrixMessageContent::Image(image_content, mime_type, media_bytes) => Some(Message {
author: Author::User,
sender_id: Some(matrix_message.sender_id.clone()),
content: MessageContent::Image(ImageDetails::new(
image_content.clone(),
mime_type.clone(),
@@ -90,5 +110,15 @@ fn convert_user_message(matrix_message: &MatrixMessage) -> Option<Message> {
)),
timestamp: matrix_message.timestamp.to_owned(),
}),
MatrixMessageContent::File(file_content, mime_type, media_bytes) => Some(Message {
author: Author::User,
sender_id: Some(matrix_message.sender_id.clone()),
content: MessageContent::File(FileDetails::new(
file_content.clone(),
mime_type.clone(),
media_bytes.clone(),
)),
timestamp: matrix_message.timestamp.to_owned(),
}),
}
}

View File

@@ -2,7 +2,9 @@ use chrono::{DateTime, Utc};
use regex::Regex;
use mxlink::matrix_sdk::ruma::OwnedUserId;
use mxlink::matrix_sdk::ruma::events::room::message::ImageMessageEventContent;
use mxlink::matrix_sdk::ruma::events::room::message::{
FileMessageEventContent, ImageMessageEventContent,
};
use mxlink::mime::Mime;
#[derive(Clone)]
@@ -18,6 +20,7 @@ pub enum MatrixMessageContent {
Text(String),
Notice(String),
Image(ImageMessageEventContent, Mime, Vec<u8>),
File(FileMessageEventContent, Mime, Vec<u8>),
}
#[derive(Clone)]

View File

@@ -119,7 +119,7 @@ async fn get_matrix_messages_in_reply_chain_native(
AnySyncMessageLikeEvent::RoomMessage(room_message) => {
if let SyncMessageLikeEvent::Original(room_message_original) = room_message {
match room_message_original.content.relates_to {
Some(Relation::Reply { in_reply_to }) => Some(in_reply_to.event_id.clone()),
Some(Relation::Reply(reply)) => Some(reply.in_reply_to.event_id.clone()),
_ => None,
}
} else {
@@ -234,6 +234,7 @@ pub async fn convert_matrix_native_event_to_matrix_message(
MessageType::Text(text_content) => (text_content.body.clone(), false),
MessageType::Notice(notice_content) => (notice_content.body.clone(), true),
MessageType::Image(image_content) => (image_content.body.clone(), false),
MessageType::File(file_content) => (file_content.body.clone(), false),
_ => return Ok(None),
};
@@ -291,6 +292,67 @@ pub async fn convert_matrix_native_event_to_matrix_message(
}));
}
if let MessageType::File(file_content) = &room_message.msgtype {
let media_request = mxlink::matrix_sdk::media::MediaRequestParameters {
source: file_content.source.to_owned(),
format: mxlink::matrix_sdk::media::MediaFormat::File,
};
let file_name = file_content
.filename
.clone()
.unwrap_or(file_content.body.clone());
let mime_type = file_content
.info
.as_ref()
.and_then(|info| info.mimetype.clone())
.and_then(|mimetype| mimetype.parse::<mxlink::mime::Mime>().ok())
.unwrap_or_else(|| get_mime_type_from_file_name(&file_name));
tracing::debug!("Determined mime type {} for file {}", mime_type, file_name);
if mime_type == mxlink::mime::APPLICATION_OCTET_STREAM {
tracing::debug!(
"Skipping file {} with unsupported MIME type {}. It will be represented as a text message.",
file_name,
mime_type,
);
return Ok(Some(MatrixMessage {
sender_id: matrix_native_event.sender().to_owned(),
content: MatrixMessageContent::Text(format!(
"[A file ({}) was attached but skipped because its content type ({}) is not supported. Let the user know.]",
file_name, mime_type,
)),
mentioned_users,
timestamp,
}));
}
let span = tracing::debug_span!("get_media_content", file_name = %file_name, mime_type = %mime_type);
let media_bytes = matrix_link
.client()
.media()
.get_media_content(&media_request, true)
.instrument(span)
.await?;
tracing::debug!(
"Downloaded {} bytes for file {}",
media_bytes.len(),
file_name
);
return Ok(Some(MatrixMessage {
sender_id: matrix_native_event.sender().to_owned(),
content: MatrixMessageContent::File(file_content.clone(), mime_type, media_bytes),
mentioned_users,
timestamp,
}));
}
Ok(Some(MatrixMessage {
sender_id: matrix_native_event.sender().to_owned(),
content: if is_notice {
@@ -358,11 +420,11 @@ pub async fn determine_interaction_context_for_room_event(
)
.await
}
Relation::Reply { in_reply_to } => {
Relation::Reply(reply) => {
determine_interaction_context_for_room_event_related_to_reply(
current_event,
current_event_is_mentioning_bot,
in_reply_to.event_id.clone(),
reply.in_reply_to.event_id.clone(),
)
.await
}

View File

@@ -1,5 +1,6 @@
use mxlink::matrix_sdk::ruma::events::room::message::{
AudioMessageEventContent, ImageMessageEventContent, MessageType, TextMessageEventContent,
AudioMessageEventContent, FileMessageEventContent, ImageMessageEventContent, MessageType,
TextMessageEventContent,
};
use mxlink::matrix_sdk::ruma::{OwnedEventId, OwnedUserId};
@@ -30,6 +31,7 @@ pub enum MessagePayload {
Text(TextMessageEventContent),
Audio(AudioMessageEventContent),
Image(ImageMessageEventContent),
File(FileMessageEventContent),
Reaction {
key: String,
@@ -57,6 +59,7 @@ impl TryInto<MessagePayload> for MessageType {
MessagePayload::Audio(audio_content)
}
MessageType::Image(image_content) => MessagePayload::Image(image_content),
MessageType::File(file_content) => MessagePayload::File(file_content),
other => {
return Err(format!("Unsupported message type: {:?}", other));
}

View File

@@ -5,8 +5,9 @@ use super::roomconfig::RoomConfig;
use crate::entity::roomconfig::{
SpeechToTextFlowType, SpeechToTextMessageTypeForNonThreadedOnlyTranscribedMessages,
TextGenerationAutoUsage, TextGenerationPrefixRequirementType, TextToSpeechBotMessagesFlowType,
TextToSpeechUserMessagesFlowType, defaults as roomconfig_defaults,
TextGenerationAutoUsage, TextGenerationPrefixRequirementType, TextGenerationSenderContextMode,
TextToSpeechBotMessagesFlowType, TextToSpeechUserMessagesFlowType,
defaults as roomconfig_defaults,
};
#[derive(Debug)]
@@ -135,6 +136,20 @@ impl RoomConfigContext {
.unwrap_or(false)
}
pub fn text_generation_sender_context_mode(&self) -> TextGenerationSenderContextMode {
self.room_config
.settings
.text_generation
.sender_context_mode
.or({
self.global_config
.fallback_room_settings
.text_generation
.sender_context_mode
})
.unwrap_or(roomconfig_defaults::TEXT_GENERATION_SENDER_CONTEXT_MODE)
}
pub fn text_generation_prefix_requirement_type(&self) -> TextGenerationPrefixRequirementType {
self.room_config
.settings

View File

@@ -1,5 +1,7 @@
use super::{SpeechToTextFlowType, SpeechToTextMessageTypeForNonThreadedOnlyTranscribedMessages};
use super::{TextGenerationAutoUsage, TextGenerationPrefixRequirementType};
use super::{
TextGenerationAutoUsage, TextGenerationPrefixRequirementType, TextGenerationSenderContextMode,
};
use super::{TextToSpeechBotMessagesFlowType, TextToSpeechUserMessagesFlowType};
pub const TEXT_GENERATION_PREFIX_REQUIREMENT_TYPE: TextGenerationPrefixRequirementType =
@@ -7,6 +9,9 @@ pub const TEXT_GENERATION_PREFIX_REQUIREMENT_TYPE: TextGenerationPrefixRequireme
pub const TEXT_GENERATION_AUTO_USAGE: TextGenerationAutoUsage = TextGenerationAutoUsage::Always;
pub const TEXT_GENERATION_SENDER_CONTEXT_MODE: TextGenerationSenderContextMode =
TextGenerationSenderContextMode::Disabled;
pub const TEXT_TO_SPEECH_BOT_MESSAGES_FLOW_TYPE: TextToSpeechBotMessagesFlowType =
TextToSpeechBotMessagesFlowType::OnDemandForVoice;

View File

@@ -16,7 +16,9 @@ pub use handler::RoomSettingsHandler;
pub use speech_to_text::{
SpeechToTextFlowType, SpeechToTextMessageTypeForNonThreadedOnlyTranscribedMessages,
};
pub use text_generation::{TextGenerationAutoUsage, TextGenerationPrefixRequirementType};
pub use text_generation::{
TextGenerationAutoUsage, TextGenerationPrefixRequirementType, TextGenerationSenderContextMode,
};
pub use text_to_speech::{TextToSpeechBotMessagesFlowType, TextToSpeechUserMessagesFlowType};
#[derive(Clone, Debug, Deserialize, Serialize, EventContent)]

View File

@@ -14,6 +14,9 @@ pub struct RoomSettingsTextGeneration {
/// When enabled, the bot will automatically tokenize messages and try to shorten the message context intelligently.
pub context_management_enabled: Option<bool>,
/// Controls how each message in the conversation context is annotated with sender metadata.
pub sender_context_mode: Option<TextGenerationSenderContextMode>,
/// Allows customizing the system prompt that the agent would use
pub prompt_override: Option<String>,
@@ -111,3 +114,46 @@ impl std::fmt::Display for TextGenerationAutoUsage {
}
}
}
#[derive(Clone, Copy, Debug, Deserialize, Serialize, PartialEq)]
pub enum TextGenerationSenderContextMode {
#[serde(rename = "disabled")]
Disabled,
#[serde(rename = "matrix_user_id")]
MatrixUserId,
#[serde(rename = "matrix_user_id_and_timestamp")]
MatrixUserIdAndTimestamp,
}
impl TextGenerationSenderContextMode {
pub fn choices() -> Vec<Self> {
vec![
Self::Disabled,
Self::MatrixUserId,
Self::MatrixUserIdAndTimestamp,
]
}
pub fn from_str(s: &str) -> Option<Self> {
match s {
"disabled" => Some(Self::Disabled),
"matrix_user_id" => Some(Self::MatrixUserId),
"matrix_user_id_and_timestamp" => Some(Self::MatrixUserIdAndTimestamp),
_ => None,
}
}
}
impl std::fmt::Display for TextGenerationSenderContextMode {
fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
match self {
TextGenerationSenderContextMode::Disabled => write!(f, "disabled"),
TextGenerationSenderContextMode::MatrixUserId => write!(f, "matrix_user_id"),
TextGenerationSenderContextMode::MatrixUserIdAndTimestamp => {
write!(f, "matrix_user_id_and_timestamp")
}
}
}
}

View File

@@ -6,8 +6,8 @@ use mxlink::helpers::account_data_config::RoomConfigManager as AccountDataRoomCo
pub use entity::{RoomConfig, RoomConfigCarrierContent, RoomSettings, RoomSettingsHandler};
pub use entity::{
SpeechToTextFlowType, SpeechToTextMessageTypeForNonThreadedOnlyTranscribedMessages,
TextGenerationAutoUsage, TextGenerationPrefixRequirementType, TextToSpeechBotMessagesFlowType,
TextToSpeechUserMessagesFlowType,
TextGenerationAutoUsage, TextGenerationPrefixRequirementType, TextGenerationSenderContextMode,
TextToSpeechBotMessagesFlowType, TextToSpeechUserMessagesFlowType,
};
pub type RoomConfigurationManager =

View File

@@ -1,3 +1,9 @@
// rustc 1.94+ trips a query-depth overflow when computing async layouts in
// the matrix-sdk timeline future graph. matrix-rust-sdk PR #6489 raises the
// limit, but `recursion_limit` is per-crate and applies to the crate currently
// being compiled — so the consumer has to repeat it.
#![recursion_limit = "256"]
mod agent;
mod bot;
mod controller;

View File

@@ -249,6 +249,13 @@ pub fn status_text_generation_entry_context_management(value: bool, set_where: &
format!("- ♻️ Context management: `{}` ({})\n", value, set_where)
}
pub fn status_text_generation_entry_sender_context(
value: impl std::fmt::Display,
set_where: &str,
) -> String {
format!("- 👤 Sender context mode: `{}` ({})\n", value, set_where)
}
pub fn status_text_generation_entry_prompt(value: &str, set_where: &str) -> String {
let value = value.trim();

View File

@@ -132,6 +132,18 @@ pub fn text_generation_context_management_intro() -> String {
)
}
pub fn text_generation_sender_context_heading() -> &'static str {
"👤 Sender Context Mode"
}
pub fn text_generation_sender_context_intro() -> String {
format!(
"{}\n{}",
"Controls whether the bot attaches sender information to conversation messages before sending them to the model.",
"`disabled` leaves messages unchanged, `matrix_user_id` adds `[sender=@alice:example.com]`, and `matrix_user_id_and_timestamp` adds `[sender=@alice:example.com sent_at=2026-03-23T14:30:00Z]`. Enabling this sends Matrix user IDs, and optionally timestamps, to the model provider.",
)
}
pub fn text_generation_prompt_override_heading() -> &'static str {
"⌨️ Prompt Override"
}

View File

@@ -17,18 +17,7 @@ pub fn get_file_extension(mime_type: &mime::Mime) -> String {
}
pub fn get_mime_type_from_file_name(file_name: &str) -> mime::Mime {
let extension = file_name.rsplit('.').next().unwrap_or("");
match extension.to_lowercase().as_str() {
"jpg" | "jpeg" => mime::IMAGE_JPEG,
"png" => mime::IMAGE_PNG,
"gif" => mime::IMAGE_GIF,
"webp" => "image/webp".parse().unwrap(),
"svg" => mime::IMAGE_SVG,
"tiff" | "tif" => "image/tiff".parse().unwrap(),
"bmp" => "image/bmp".parse().unwrap(),
"heic" | "heif" => "image/heic".parse().unwrap(),
"avif" => "image/avif".parse().unwrap(),
_ => mime::APPLICATION_OCTET_STREAM,
}
mime_guess::from_path(file_name)
.first()
.unwrap_or(mime::APPLICATION_OCTET_STREAM)
}