* refactor: merge desktop/ into frontend/, eliminate duplicate Tauri scaffolding
The project had two overlapping directories: desktop/ (Tauri Rust backend +
stale React components) and frontend/ (real React app + dead Tauri stub).
This consolidates everything under frontend/:
- Move desktop/src-tauri/ → frontend/src-tauri/ (the real 1,720-line Rust
backend with Ollama sidecar, backend lifecycle, cloud keys, overlay, etc.)
- Preserve 9 old desktop React components in frontend/src/components/Desktop/
(excluded from TS build — APIs have drifted, kept for future integration)
- Delete the old frontend/src-tauri/ stub (246 lines, never compiled)
- Delete desktop/ entirely
- Fix tauri.conf.json frontendDist path (../../frontend/dist → ../dist)
- Update CI workflow, bump script, .gitignore, and docs paths
- Rename setup/ → Setup/ for consistent PascalCase component directories
* feat: add memory UI, settings, and fix memory API routes
- Add Memory tab to Data Sources page with stats, search, index path,
and manual store functionality
- Add Memory section to Settings page with backend picker, context
injection toggle, and parameter sliders (top_k, min_score, max_tokens)
- Add memory API functions to frontend (getMemoryStats, searchMemory,
storeMemory, indexMemoryPath, getMemoryConfig)
- Fix backend /v1/memory/* routes to use app-level memory backend
instead of creating fresh SQLiteMemory instances per request
- Add GET /v1/memory/config and POST /v1/memory/index endpoints
- Gracefully handle missing Rust backend (return defaults instead of 500)
- Fix nested scroll in Agents Interact tab (use viewport-relative height)
* fix: memory API routes, dialog plugin, and UI polish
- Fix memory API: use backend.retrieve() not .search(), .count() not
.stats() to match actual SQLiteMemory interface
- Handle None backend gracefully in index endpoint (503 instead of crash)
- Expand ~ in index path (expanduser + resolve)
- Install @tauri-apps/plugin-dialog for native folder picker in Tauri
- Browse button only shows in Tauri (browser can't get absolute paths)
- Redesign Memory tab: proper cards, color-coded search scores, two-column
layout for index/store, loading spinners, accent gradient on stats card
---------
Co-authored-by: Jon Saad-Falcon <41205309+jonsaadfalcon@users.noreply.github.com>
* feat: add Apple Contacts connector (macOS AddressBook)
Reads directly from ~/Library/Application Support/AddressBook/AddressBook-v22.abcddb
in read-only mode. Extracts names, phone numbers, emails, postal addresses, URLs,
social profiles, and notes for each contact. Requires Full Disk Access like
iMessage and Apple Notes connectors.
- New connector: src/openjarvis/connectors/apple_contacts.py
- Registered in connectors/__init__.py
- Added frontend catalog entry (PIM category) and icon mapping
- MCP tools: contacts_search, contacts_get_contact
* test: add comprehensive tests for Apple Contacts connector
15 tests covering: is_connected (exists/missing), sync yields all real
contacts (skips system rows), extracts all field types (phone, email,
address, URL, social, notes), cleans Apple label markup, org-only
contacts, minimal contacts, since filter, sync_status tracking,
structured metadata, disconnect, mcp_tools, registry, empty DB,
missing DB. Uses fake SQLite database — no real Contacts DB needed.
* fix: scan iCloud/Exchange source databases for all contacts
The main AddressBook database only contains locally-created contacts.
Synced contacts (iCloud, Exchange, etc.) live under
Sources/<UUID>/AddressBook-v22.abcddb. Now scans all source databases
and deduplicates by ZUNIQUEID across sources.
Adds tests for multi-source scanning and cross-source deduplication.
The project had two overlapping directories: desktop/ (Tauri Rust backend +
stale React components) and frontend/ (real React app + dead Tauri stub).
This consolidates everything under frontend/:
- Move desktop/src-tauri/ → frontend/src-tauri/ (the real 1,720-line Rust
backend with Ollama sidecar, backend lifecycle, cloud keys, overlay, etc.)
- Preserve 9 old desktop React components in frontend/src/components/Desktop/
(excluded from TS build — APIs have drifted, kept for future integration)
- Delete the old frontend/src-tauri/ stub (246 lines, never compiled)
- Delete desktop/ entirely
- Fix tauri.conf.json frontendDist path (../../frontend/dist → ../dist)
- Update CI workflow, bump script, .gitignore, and docs paths
- Rename setup/ → Setup/ for consistent PascalCase component directories
Update all human-readable strings (docstrings, comments, descriptions,
log messages, TOML config descriptions) to correctly say
"DeepResearchBench" instead of "LiveResearchBench" for the
deep_research_bench benchmark. Class names and file paths are
unchanged for backward compatibility.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Add Salesforce LiveResearchBench (liveresearchbench) as a new benchmark:
- Dataset provider loading 80 expert-curated research tasks with 543
checklist items from HuggingFace (Salesforce/LiveResearchBench)
- Checklist-based scorer evaluating coverage + quality dimensions
- Groups multiple checklist rows per question into single EvalRecords
Also add "deepresearch" as a CLI alias for the existing DeepResearchBench
benchmark (previously only available as "liveresearch").
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Ahead of the 2026-06-02 forced-Node-24 deadline, upgrade the remaining
javascript actions flagged in the workflow annotations:
- actions/setup-python v5 → v6
- actions/setup-node v4 → v6 (skipping v5)
- actions/cache v4 → v5
- actions/deploy-pages v4 → v5
- actions/upload-pages-artifact v3 → v5.0.0
upload-pages-artifact has no rolling @v5 major tag — only the specific
v5.0.0 release — so it gets an explicit pin, same as astral-sh/setup-uv.
Not touched: swatinem/rust-cache@v2 (still on v2 majors, not flagged),
dtolnay/rust-toolchain@stable (composite), tauri-apps/tauri-action@v0
(composite), anthropics/claude-code-action@v1 (still latest major).
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
astral-sh/setup-uv does not publish a rolling @v8 major tag — only the
specific v8.0.0 release. Previous commit's @v8 reference failed to
resolve and broke CI/Docs/PyPI-publish runs.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- openjarvis.__version__ now comes from importlib.metadata.version("openjarvis")
instead of a hardcoded string, so it stays in sync with pyproject.toml on every
release. Tests assert against openjarvis.__version__ rather than a literal.
- Bump actions/checkout@v4 → @v6 and astral-sh/setup-uv@v4 → @v8 across every
workflow ahead of the 2026-06-02 Node.js 20 deprecation on GitHub Actions runners.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Now that the `openjarvis` name is available on PyPI, rename the
distribution from `OpenJarvisAI` to `OpenJarvis` so `pip install
openjarvis` matches the import name and the in-code install hints
(e.g. `pip install openjarvis[channel-gmail]`).
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix(cloud-router): prevent local HF org models from being misrouted to OpenRouter
get_provider() previously routed any "/" model to OpenRouter, which caused
locally-served models (e.g. mlx-community/Qwen3.5-27B-4bit-DWQ) to be sent
to the cloud instead of localhost:8080.
Adds _LOCAL_HF_ORGS module-level allowlist covering mlx-community/,
bartowski/, unsloth/, and lmstudio-community/. Returns None early for
these before the openrouter "/" check.
* fix(doctor): respect preferred_engine/engine.default when reporting default model
_check_default_model() iterated engines in sorted alphabetical order,
causing engines like 'lemonade' to be reported as the host for the
default model even when 'mlx' was configured as preferred_engine and
engine.default. Now checks the configured engine first, then falls back
to alphabetical scan.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
---------
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
The clone tracking workflow needs to push directly to main, but branch
protection rules block GITHUB_TOKEN pushes. Using the PAT for checkout
allows the push to succeed.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
AppleScript's `tell application "X"` is an implicit launch directive —
macOS starts the app if it is not running. Both scripts in apple_music.py
used this pattern unconditionally, causing Music.app to open on every
connector status poll (frontend load, SyncScheduler ticks) for all macOS
users regardless of whether they use the Apple Music connector.
Guard both scripts with `application "Music" is running` so that
is_connected() returns False gracefully when Music is closed, without
ever launching it unsolicited.
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
get_provider() previously routed any "/" model to OpenRouter, which caused
locally-served models (e.g. mlx-community/Qwen3.5-27B-4bit-DWQ) to be sent
to the cloud instead of localhost:8080.
Adds _LOCAL_HF_ORGS module-level allowlist covering mlx-community/,
bartowski/, unsloth/, and lmstudio-community/. Returns None early for
these before the openrouter "/" check.
- Move download/clone/star badges from docs/index.md to README.md
- Use TRAFFIC_TOKEN secret instead of GITHUB_TOKEN for Traffic API
(requires admin-level access that GITHUB_TOKEN doesn't provide)
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
openjarvis was already taken on PyPI, so the package is registered as
OpenJarvisAI. Update pyproject.toml and docs badges to match.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Per-config max_turns:
- Add `max_turns` field to ExecutionConfig and RunConfig dataclasses
- Parse `[run] max_turns` from eval TOML configs
- Plumb through `_build_backend` to JarvisAgentBackend, which sets
`builder._config.agent.max_turns` before building the system
- Default `None` falls back to `JarvisConfig.agent.max_turns` (10) so
existing configs are unaffected
Why: Trinity-Large is a "thinking" model that consumes agent-loop turns
producing intermediate reasoning, not just tool calls. With the default
max_turns=10 it hit the cap on 25/50 GAIA tasks (50%) and 39/50
LiveResearchBench tasks (78%) before producing a final answer. With
max_turns=50 (set per-config in this commit's gaia-trinity-large.toml
and liveresearch-trinity-large.toml updates), Trinity LR jumped from
12.0% to 72.0% — equal to Qwen-27B. Without this field the only
workaround was a runtime `OPENJARVIS_CONFIG` TOML hack.
Misc config bumps:
- `gaia-qwen-2b.toml`, `liveresearch-qwen-2b.toml`,
`taubench-telecom-qwen2b.toml`: max_workers 1 -> 8 to use the
parallel runner from #207
- `gaia-trinity-large.toml`, `liveresearch-trinity-large.toml`:
add `max_turns = 50` (using the new field)
Co-authored-by: Jon Saad-Falcon <jonsaadfalcon@gpu-dp-vg94q-lccz2.t-f6a9b26f-56c0-4f98-88fc-705e5d9b1714.svc.cluster.local>
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Qwen3.5 MoE models provide better quality per GB of memory since only
a fraction of parameters are active per token. Updates the tier table:
≤8 GB → qwen3.5:2b
8-16GB → qwen3.5:4b
16-32GB → qwen3.5:9b
32GB+ → qwen3.5:27b
Closes#132 — jarvis init now recommends a Qwen3 dense model based on
available memory using an explicit tier table:
≤8 GB → qwen3:1.7b (~1.1 GB)
8-16GB → qwen3:4b (~2.5 GB)
16-32GB → qwen3:8b (~5.2 GB)
32GB+ → qwen3:14b (~9.3 GB)
Adds qwen3:0.6b, 1.7b, 4b, 14b, 30b to the model catalog. Falls back
to scanning all compatible models if the tiered pick doesn't support
the selected engine. Shows the user why the model was chosen and how
to download later if declined.
Closes#182 — the claude-code-action attempts OIDC token exchange when
no github_token is provided, which fails without the Anthropic GitHub
App installed. Passing secrets.GITHUB_TOKEN skips the OIDC flow.
Closes#179 — traces were never persisted in the serve path because:
1. TraceStore.subscribe_to_bus() was never called in app.py
2. AgentExecutor was constructed without trace_store in serve.py and
agent_manager_routes.py
3. The /v1/traces API returned raw dataclass fields instead of the
frontend-expected format
Fixes all three: wires bus subscription, passes trace_store to all
executor instantiation sites, and adds _serialise_trace() for proper
API response formatting.
Closes#164 — Qwen and other models with baked-in corporate identities
would claim to be cloud services when running locally. Adds a
configurable default_system_prompt to AgentConfig that is injected as a
fallback when no other system prompt is provided. The default tells the
model it is running locally through OpenJarvis. Users can override or
clear it via config.toml.
