AI-Assistant

Bridge the ComfyUI anime image pipeline into the chat SSE stream so image generation renders inline in the chat bubble. Use when: changing how image-generation events reach the chat UI, editing the force_image_bridge gate or the SSE forwarder in stream.py, touching cancel / GPU-leak handling (GeneratorExit → request_cancel → /free), the error_class taxonomy, the Phase 1b thinking-block nesting, or the Phase 3 WS live-preview; debugging why an image turn stalls, double-renders, leaks VRAM, or shows the wrong error; or running the PR4 verification harness. This subsystem deliberately spans services/chatbot/ AND app/image_pipeline/.

Choose which repository skill to activate for a given task. Use when: starting any chatbot, MCP, config, or docs task to determine the best skill or skill combination; reviewing whether the right skills were applied; or deciding whether to combine multiple skills for a cross-cutting change.

Repository maintenance and coding-agent instructions for AI-Assistant.

Reason about the impact of chatbot/core changes on GitHub workflows, tests, security scans, and CI behavior. Use when: adding or changing routes, tools, providers, or dependencies; reviewing a PR that might affect CI; debugging CI failures after a chatbot change; checking whether a code change needs a workflow or docs update; or auditing workflow assumptions after startup or env changes.

Map changed files to the smallest sufficient verification set. Use when: deciding which tests to run after a code change, explaining why a test scope is enough, or choosing between unit tests, integration tests, and manual checks.

Use when the user needs to run GitNexus CLI commands like analyze/index a repo, check status, clean the index, generate a wiki, or list indexed repos. Examples: "Index this repo", "Reanalyze the codebase", "Generate a wiki"

Keep documentation aligned with runtime behavior. Use when: changing ports, entry points, startup commands, env variables, service responsibilities, or removing/renaming scripts. Requires a runtime-vs-docs comparison before finalizing any change that alters observable behavior.

Work safely with provider configuration, API keys, model routing, and environment dependencies for the core chatbot. Use when: adding or changing an LLM provider, modifying API key loading, changing model defaults or fallback chains, debugging missing-key errors, reviewing config drift across .env files, or checking whether a provider change affects tools or docs.

Guide changes to chatbot search tools, reverse-image flows, fallback logic, and auto-trigger behavior. Use when: adding or modifying a search tool, changing fallback order, editing auto-search keyword lists, wiring a new tool button in the UI, changing response formatting for search results, debugging a tool that stopped returning results, or reviewing search cascade behavior.

Reason about chatbot thinking modes and route requests to the correct mode. Use when adding or changing a thinking mode, modifying mode selection UI, changing how mode affects backend routing or tool dispatch, debugging broken thinking/reasoning display, or reviewing SSE event flow for thinking modes.

Get broader context and a higher-level perspective before editing unfamiliar code. Use when entering an unknown module, when a change feels risky but scope is unclear, or when asked to refactor a section without a clear plan.

Keep the chat UI synchronized with backend routes, tool selectors, mode selectors, and response rendering for the core chatbot. Use when: adding or changing a UI control that calls a backend route, modifying a tool or mode selector, changing a route that the frontend depends on, debugging broken rendering or missing UI state, reviewing whether a functional change needs a UI update, or checking that the frontend and backend agree on payload field names.

Audit and safely edit the core chatbot request path: route registration, blueprint wiring, provider routing, tool dispatch, SSE streaming, and UI-to-backend wiring. Use when adding or changing a chat route, modifying provider selection, wiring a new tool into the UI, debugging broken streaming, or reviewing blueprint registration.

Disciplined 6-phase debug loop for hard bugs and performance regressions. Use when a bug is reported, something is broken/throwing/failing, or a performance regression is described.

Improve diagnostics for the core chatbot and tools without adding noisy or unsafe logging. Use when: adding or reviewing log statements, debugging silent failures, auditing error handling, checking for secret leakage in logs, improving startup or provider failure signals, or deciding what log level to use for a new code path.

Debug startup failures, port drift, and service health for the core chatbot stack. Use when: a service fails to start, a health check returns unexpected results, ports or entry points are mismatched across scripts and docs, dependency profiles need verification, Docker or CI startup assumptions need checking, or runtime behavior no longer matches documentation.

Audit and enforce the shared environment loading contract across all services. Use when: adding, renaming, or removing an environment variable; changing config loading logic; debugging missing env values at runtime; adding a new service or dependency profile; reviewing whether a change affects chatbot only, MCP only, or multiple services; or checking for hardcoded ports, URLs, or secrets.

Define and enforce response-shape contracts for every tool function, route handler, SSE event, and MCP tool in the chatbot stack. Use when: adding or modifying a tool function return type, changing a route's JSON response, adding or changing an SSE event payload, modifying an MCP tool's return schema, debugging frontend crashes from missing response fields, reviewing backward compatibility of response changes, or tracing how a response shape flows from backend to frontend.

Maintain the character picker, character registry, and local job queue tier added to AI-Assistant. Use when: editing storage/character_db/ data files; modifying core/character_registry.py or core/job_queue.py; changing /api/characters/* or /api/jobs/* routes; touching the character-picker.js / job-queue-panel.js frontend modules; integrating character_key into a new image-gen flow; or extending the queue lifecycle with new states.

Test-driven development with red-green-refactor loop. Use when building a new feature or fixing a bug with TDD, or when asked for test-first development.

Guide changes to the MCP server: adding or modifying tools, resources, and prompts. Use when: adding a new MCP tool, resource, or prompt; changing an existing tool's input/output contract; debugging tool registration or response shape; reviewing MCP server code; checking chatbot-side MCP route compatibility; or updating MCP documentation.

Teach Copilot how to choose the correct dependency profile when working on the core chatbot and tools, and avoid dragging image-service dependencies into chatbot-only tasks. Use when: adding, removing, or upgrading a Python package; reviewing a dependency change; debugging import errors or version conflicts; checking whether a package belongs in venv-core or venv-image; auditing CI workflow install steps; or verifying that a chatbot-only change does not pull in image-stack packages.