The llamacpp provider was instantiated with vision disabled by default, causing image transfers from Telegram to fail. Use new_with_vision() with vision enabled, matching the behavior of other compatible providers.
Fixes#3802
The deferred MCP tools section in the system prompt only listed tool
names inside <available-deferred-tools> tags without any instruction
telling the LLM to call tool_search to activate them. In daemon and
Telegram mode, where conversations are shorter and less guided, the
LLM never discovered it should call tool_search, so deferred tools
were effectively unavailable.
Add a "## Deferred Tools" heading with explicit instructions that
the LLM MUST call tool_search before using any listed tool. This
ensures the LLM knows to activate deferred tools in all modes
(CLI, daemon, Telegram) consistently.
Also add tests covering:
- Instruction presence in the deferred section
- Multiple-server deferred tool search
- Cross-server keyword search ranking
- Activation persistence across multiple tool_search calls
- Idempotent re-activation
When a provider returns a context-size-exceeded error, truncate the
oldest non-system messages from conversation history and retry instead
of immediately bailing out. This enables local models with small
context windows (llamafile, llama.cpp) to work by automatically
fitting the conversation within available context.
Closes#3894
Add support for switching AI models at runtime during a conversation.
The model_switch tool allows users to:
- Get current model state
- List available providers
- List models for a provider
- Switch to a different model
The switch takes effect immediately for the current conversation by
recreating the provider with the new model after tool execution.
Risk: Medium - internal state changes and provider recreation
* fix(providers): preserve conversation context in Claude Code CLI provider
Override chat_with_history to format full multi-turn conversation
history into a single prompt for the claude CLI, instead of only
forwarding the last user message.
Closes#3878
* fix(providers): fix ETXTBSY race in claude_code tests
Use OnceLock to initialize the fake_claude.sh test script exactly
once, preventing "Text file busy" errors when parallel tests
concurrently write and execute the same script file.
Handle Schedule::At jobs in reschedule_after_run by disabling them
instead of rescheduling to a past timestamp. Also add a fallback in
persist_job_result to disable one-shot jobs if removal fails.
Closes#3868
When using Channel mode with dynamic classification and routing, the
route-specific `api_key` from `[[model_routes]]` was silently dropped.
The system always fell back to the global `api_key`, causing 401 errors
when routing to `custom:` providers that require distinct credentials.
Root cause: `ChannelRouteSelection` only stored provider + model, and
`get_or_create_provider` always used `ctx.api_key` (the global key).
Changes:
- Add `api_key` field to `ChannelRouteSelection` so the matched route's
credential survives through to provider creation.
- Update `get_or_create_provider` to accept and prefer a route-specific
`api_key` over the global key.
- Use a composite cache key (provider name + api_key hash) to prevent
cache poisoning when multiple routes target the same provider with
different credentials.
- Wire the route api_key through query classification matching and the
`/model` (SetModel) command path.
Fixes#3838
The Dockerfile and Dockerfile.debian COPY `firmware/`, `crates/robot-kit/`,
and `crates/robot-kit/Cargo.toml`, but `.dockerignore` excludes both
`firmware/` and `crates/robot-kit/`, causing COPY failures during build.
Since these are hardware-only paths not needed for the Docker runtime:
- Remove COPY commands for `firmware/` and `crates/robot-kit/`
- Remove dummy `crates/robot-kit/src` creation in dep-caching steps
- Use sed to strip `crates/robot-kit` from workspace members in the
copied Cargo.toml so Cargo doesn't look for the missing manifest
Fixes#3836
Fetch the current pairing code from GET /admin/paircode (localhost-only)
and display it in both the initial PairingDialog and the /pairing
management page. Users no longer need to check the terminal to find
the 6-digit code — it appears directly in the web UI.
Falls back gracefully when the admin endpoint is unreachable (e.g.
non-localhost access), showing the original "check your terminal" prompt.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Update Facebook group link from /groups/zeroclaw to /groups/zeroclawlabs
across all 31 README locale files. Add Discord, TikTok, and RedNote
social badges to the badge section of all READMEs.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
cargo-audit uses .cargo/audit.toml (not deny.toml) for its ignore
list. These 3 wasmtime advisories are transitive via extism 1.13.0
with no upstream fix available. Plugin system is feature-gated.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
RUSTSEC-2026-0006, RUSTSEC-2026-0020, RUSTSEC-2026-0021 are all in
wasmtime 37.x pinned by extism. No newer extism release available.
Plugin system is behind a feature flag to limit exposure.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add cargo-deny ignore entries for RUSTSEC-2024-0388 (derivative),
RUSTSEC-2025-0057 (fxhash), and RUSTSEC-2025-0119 (number_prefix).
All are transitive dependencies we cannot directly control.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Sync Cargo.lock with new Extism/WASM plugin dependencies and apply
rustfmt line-wrap fix in gateway WebSocket handler.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Wire WASM plugin tools into all_tools_with_runtime() behind
cfg(feature = "plugins-wasm"), discovering and registering tool-capable
plugins from the configured plugins directory at startup.
- Add /api/plugins gateway endpoint (cfg-gated) for listing plugin status.
- Add mod plugins declaration to main.rs binary crate so crate::plugins
resolves when the feature is enabled.
- Add unit tests for PluginHost: empty dir, manifest discovery, capability
filtering, lookup, and removal.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add a standalone example plugin demonstrating the WASM plugin interface:
- example-plugin/Cargo.toml: cdylib crate targeting wasm32-wasip1
- example-plugin/src/lib.rs: mock weather tool using extism-pdk
- example-plugin/manifest.toml: plugin manifest declaring tool capability
This crate is intentionally NOT added to the workspace members since it
targets wasm32-wasip1 and would break the main build.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Introduce the WASM plugin system foundation:
- Add extism 1.9 as an optional dependency behind `plugins-wasm` feature
- Create `src/plugins/` module with manifest types, error types, and stub host
- Add `Plugin` CLI subcommands (list, install, remove, info) behind cfg gate
- Add `PluginsConfig` to the config schema with sensible defaults
All plugin code is behind `#[cfg(feature = "plugins-wasm")]` so the default
build is unaffected.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add device_registry, pending_pairings to test AppState instances and
pairing_dashboard to test GatewayConfig to fix compilation of tests
after the new pairing dashboard fields were introduced.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add Pairing page with device list table, pairing code generation,
and device revocation. Create useDevices hook for reusable device
fetching. Wire /pairing route into App.tsx router.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add ConnectParams struct for an optional first-frame connect handshake.
If the first WebSocket message is {"type":"connect",...}, connection
parameters (session_id, device_name, capabilities) are extracted and
a "connected" ack is sent back. Old clients sending "message" first
still work unchanged (backward-compatible).
Extract process_chat_message() helper to avoid duplication between
fallback first-message handling and the main message loop.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Introduce DeviceRegistry, PairingStore, and five new API endpoints:
- POST /api/pairing/initiate — generate a new pairing code
- POST /api/pairing/submit — submit code with device metadata
- GET /api/devices — list paired devices
- DELETE /api/devices/{id} — revoke a paired device
- POST /api/devices/{id}/rotate — rotate a device token
Wire into AppState and gateway router. Registry is only created
when require_pairing is enabled.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add PairingDashboardConfig struct with configurable code_length,
ttl_secs, max_pending, max_attempts, and lockout_secs fields.
Nested under GatewayConfig as `pairing_dashboard` with serde defaults.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Export commands module from lib.rs (pub mod commands) for external consumers
- Add --force and --version flags to the Update CLI command
- Wire version parameter through to check() and run() in update.rs,
supporting targeted version fetches via GitHub releases/tags API
- Add WebSocket handshake check (check_websocket_handshake) to the full
self-test suite in self_test.rs
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add `zeroclaw update` command with a 6-phase self-update pipeline:
1. Preflight — check GitHub releases API for newer version
2. Download — fetch platform-specific binary to temp dir
3. Backup — copy current binary to .bak for rollback
4. Validate — size check + --version smoke test on download
5. Swap — overwrite current binary with new version
6. Smoke test — verify updated binary runs, rollback on failure
Supports --check flag for update-check-only mode without installing.
Includes version comparison logic with unit tests.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add `zeroclaw self-test` command with two modes:
- Quick mode (--quick): 8 offline checks including config, workspace,
SQLite, provider/tool/channel registries, security policy, and version
- Full mode (default): adds gateway health and memory round-trip checks
Creates src/commands/ module structure with self_test and update stubs.
Adds indicatif and tempfile runtime dependencies for the update pipeline.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The Debian compatibility image was building from source with QEMU
cross-compilation for ARM64, which is extremely slow and was getting
cancelled by the concurrency group. Switch to using pre-built binaries
(same as the distroless image) with a debian:bookworm-slim runtime base.
- Add Dockerfile.debian.ci (mirrors Dockerfile.ci with Debian runtime)
- Update release-beta-on-push.yml to use docker-ctx + pre-built bins
- Update release-stable-manual.yml with same fix
- Drop GHA cache layers (no longer building from source)
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
Gateway WebSocket chat sessions were in-memory only — conversation
history was lost on gateway restart, macOS sleep/wake, or client
reconnect. This wires up the existing SessionBackend (SQLite) to
the gateway WS handler so sessions survive restarts and reconnections.
Changes:
- Add delete_session() to SessionBackend trait + SQLite implementation
- Add session_persistence and session_ttl_hours to GatewayConfig
- Add Agent::seed_history() to hydrate agent from persisted messages
- Initialize SqliteSessionBackend in run_gateway() when enabled
- Send session_start message on WS connect with session_id + resumed
- Persist user/assistant messages after each turn
- Add GET /api/sessions and DELETE /api/sessions/{id} REST endpoints
- Bump version to 0.5.0
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
The Turkish (tr) locale section had a duplicate "Dashboard specific
labels" block that repeated 19 keys already defined earlier, causing
TypeScript error TS1117. Moved the unique keys (provider_model,
paired_yes, etc.) into the primary dashboard section and removed
the duplicate block.
Fixes build failure introduced by #3777.
Remove tweet job from beta workflow. Update tweet-release.yml to diff
against previous stable tag (excluding betas) to capture all features
across the full release cycle. Simplify tweet format to feature-focused
style without contributor counts.
Supersedes #3575.
- Add comprehensive Simplified Chinese (zh) translations for all UI strings
- Extend and complete Turkish (tr) translations
- Fill in missing English (en) translation keys
- Reset default locale to 'en'
- Update language toggle to cycle through all three locales: en → zh → tr
Revert "always generate pairing code" to tighter security posture:
codes are only generated on first startup when no tokens exist. Add
a CLI hint to the gateway banner so operators know how to re-pair
on demand. Fix install.sh to not use --new on fresh install (avoids
invalidating the auto-generated code). Fix onboard to show an
informational message instead of a throwaway PairingGuard.
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
* feat(runtime): add configurable reasoning effort
* fix(test): add missing reasoning_effort field in live test
Add reasoning_effort: None to ProviderRuntimeOptions construction in
openai_codex_vision_e2e.rs to fix E0063 compile error.
---------
Co-authored-by: Alix-007 <267018309+Alix-007@users.noreply.github.com>
