Open Issues Need Help
View All on GitHub `gaia agent test --live` reports "All live gates passed" with no LLM backend reachable about 2 hours ago
good first issue p2 weekly-audit
Agent UI 'Delete all chats' and memory toggle settings have no tests about 2 hours ago
good first issue p2 weekly-audit
Flagship agent SPEC.md omits GAIA_ALLOWED_PATHS and the 150-step default about 2 hours ago
good first issue p3 weekly-audit
Documented mail-grant examples fail or grant an agent with no mail tools (4 places) about 2 hours ago
good first issue p2 weekly-audit
Any document Q&A startup failure is reported as 'RAG not installed' and never retried about 2 hours ago
good first issue p2 weekly-audit
The Agent UI's Download model button sends a pull that Lemonade rejects for custom models, and nothing tests it about 2 hours ago
good first issue p2 weekly-audit
fix(rag): RAGSDK.query carries earlier questions' retrieved context into later answers about 4 hours ago
bug good first issue p2 rag
A bare gaia eval agent fails pre-flight without GAIA_MEMORY_ADMIN=1 / GAIA_MEMORY_MCP_ALWAYS=1 on the backend, neither documented about 6 hours ago
good first issue p2 weekly-audit
All three CI workflows install only ".[eval]", which does not ship fastapi/uvicorn, so their own Start Agent UI backend step crashes about 6 hours ago
good first issue p2 weekly-audit
The new `reconnecting` status phase is missing from the flagship's integrator docs about 6 hours ago
good first issue p3 weekly-audit
fix(macos): the menu-bar icon is a plain square, not the GAIA logo about 6 hours ago
bug good first issue p3
ux(ui): document answers cite the upload's internal ID instead of the attached file's name about 6 hours ago
bug good first issue p2 agent-ui
good first issue p2 weekly-audit
Code index retries failed embeddings on its own, then falls back to one call per chunk about 6 hours ago
bug good first issue p2
Agent UI shell: download progress polls at 5 s instead of 2 s, and the chat-create error toast never clears about 6 hours ago
bug good first issue p3 agent-ui
Agent UI: Document Library, File Browser and Mobile access dialogs ignore Esc and don't take focus about 6 hours ago
bug good first issue p3 agent-ui
enhancement good first issue p3
fix(tui): an installed agent with a corrupt install record vanishes from the list without a warning about 6 hours ago
bug good first issue p3
Cloud models from other providers get llama.cpp-only fields until the catalog is read about 6 hours ago
bug good first issue llm p3
Small cleanups in the agent base and tool mixins (stale docstring, dead variable, registry gap, history comments) about 6 hours ago
documentation good first issue p3
Agent UI and memory guide still promise '100% local' when chat uses a cloud model (4 places) about 6 hours ago
good first issue p2 weekly-audit
ci: workflows still wired to deleted or disabled eval workflows about 6 hours ago
good first issue p2
good first issue p3
Release publisher's 409 handling decides pass/fail and has no tests (2 publishers) about 6 hours ago
good first issue p2 weekly-audit
chore(website): delete the code panel no page renders about 6 hours ago
enhancement good first issue p3
The Python syntax gate that stops agents corrupting source files is untested — its tests mock methods the code no longer calls about 6 hours ago
good first issue p2 weekly-audit
The flagship's published docs were never updated for v0.2.0 — version, tool count and hash policy all still describe 0.1.1 about 6 hours ago
good first issue p2 weekly-audit
Three of five smoke-test handshake failures still discard the child's stderr, leaving release failures with no cause about 6 hours ago
good first issue p2 weekly-audit
Every reference link in the published email agent's docs 404s — they point at a tag that was never created (11 links) about 6 hours ago
good first issue p2 weekly-audit
The code-index guide states no minimum Lemonade version, so an old server hangs for 12 minutes then fails about 6 hours ago
good first issue p3 weekly-audit
`gaia memory bootstrap --infer` stores nothing, dumps a raw truncated LLM response, and exits 0 about 6 hours ago
good first issue p2 weekly-audit
The corpus "validate the manifest" step validates nothing — it passes on the exact mistake it exists to catch about 6 hours ago
good first issue p3 weekly-audit
The AI-provider controls the guide documents unconditionally only exist on the flagship agent about 6 hours ago
good first issue p3 weekly-audit
The documented `uv run` commands build a bare `.venv` in the checkout and then fail (2 guides) about 6 hours ago
good first issue p2 weekly-audit
The flagship agent npm docs promise "No Python to install", then require the Python CLI about 6 hours ago
good first issue p2 weekly-audit
No frontend test checks that mutating requests carry `x-gaia-ui`, the header the backend now requires about 6 hours ago
good first issue p2 weekly-audit
`gaia talk --index` silently skips indexing after the guide's own install step, because the [talk] extra ships no RAG dependencies about 6 hours ago
good first issue p2 weekly-audit
`gaia agent status` is documented as showing the installed version, but the VERSION column prints `-` even for an installed, healthy agent about 6 hours ago
good first issue p3 weekly-audit
--save-baseline is documented with no warning that it overwrites the git-tracked eval/results/baseline.json, even from an all-failed run about 6 hours ago
good first issue p3 weekly-audit
The flagship copy-paste example names a fact id that does not exist in the corpus manifest about 6 hours ago
good first issue p3 weekly-audit
The agent class printed in section 3 cannot be instantiated - section 5's first line raises TypeError about 6 hours ago
good first issue p2 weekly-audit
test(connectors): OAuth refresh body encoding is asserted nowhere about 6 hours ago
good first issue p1 tests robustness
The Windows tab tells readers to run `nvm alias default 20.20.0`, but NVM for Windows has no `alias` command about 6 hours ago
good first issue p3 weekly-audit
gaia cache status reports only the Context7 cache and prints no sizes, contradicting size and location of each cache directory about 6 hours ago
good first issue p3 weekly-audit
'one of six categories' contradicts the same doc and the categories actually written to the DB about 6 hours ago
good first issue p3 weekly-audit
good first issue p3 weekly-audit
The documented eval command fails pre-flight - it needs the Agent UI backend on port 4200 and the doc never says so about 6 hours ago
good first issue p2 weekly-audit
ci(email): nothing checks that the Python and npm contract versions agree — a split ships green about 6 hours ago
good first issue p2
Wire the followups email eval into CI (report script + workflow step) about 6 hours ago
good first issue agent eval p3
gaia agent init has no --tools flag, so the CLI can't scaffold an agent with a toolset about 8 hours ago
enhancement good first issue p2 cli
documentation good first issue p1 website domain:surfaces
bug documentation good first issue p1 onboarding domain:platform
The documented compare command points at eval/results/latest/, a path the runner never creates 4 days ago
good first issue weekly-audit
good first issue weekly-audit
good first issue weekly-audit
good first issue weekly-audit
good first issue weekly-audit
good first issue weekly-audit
Tag filtering can never match anything: not one of the 91 shipped scenarios declares a tags field 5 days ago
good first issue weekly-audit
bug good first issue tests
The fork command fails as written because cp will not create the missing ~/my-skills parent 5 days ago
good first issue weekly-audit
good first issue weekly-audit
enhancement good first issue eval p3
good first issue weekly-audit
bug good first issue p2
good first issue
good first issue weekly-audit
bug good first issue lemonade 🍋 p3
bug good first issue eval p3
good first issue triage
bug good first issue p1
fix(tools): search_file can't find a file you name exactly if its extension isn't on the allowlist 9 days ago
bug good first issue p1
good first issue
good first issue
good first issue p1
bug good first issue p3 agent-ui
bug good first issue lemonade 🍋 p3
bug good first issue p3
enhancement good first issue tech debt agent-ui domain:surfaces
bug good first issue audio p3
gaia mcp commands exit 0 when they fail 10 days ago
bug good first issue p1 mcp
documentation good first issue
documentation good first issue p1 cli domain:platform
good first issue weekly-audit
bug good first issue p2
good first issue weekly-audit
bug good first issue
good first issue weekly-audit
Every stateful-agent example in the email agent's SKILL.md fails with 401 — no session token is sent 10 days ago
good first issue weekly-audit
good first issue weekly-audit
docs(cli): gaia email help says Gmail-only; cli.mdx contradicts itself on --base-url and Kill 10 days ago
good first issue
good first issue
docs: bare amd-gaia.ai links outside src/gaia 404 (issue template, install-ui scripts, CLAUDE.md) 10 days ago
good first issue
documentation good first issue p3
RAG: VLM image blocks can be split across chunks when section-based splitting path is used 10 days ago
bug good first issue p2 rag robustness domain:platform track:platform
test: skill-library tool mixin has zero dedicated tests, including the audit-gate refusal 15 days ago
good first issue
documentation good first issue p1 domain:integrations connectors
good first issue
documentation good first issue
bug good first issue p2