List all AI models available across every configured provider (Ollama, OpenAI, Anthropic, X.AI Grok (API key), vLLM, TRT-LLM, SGLang, plus subscription-CLI providers: Claude, ChatGPT/Codex, Grok). Use the returned model IDs when calling configure_council.

Update the council configuration: select which models form the council, choose a judge model, set the response mode (individual / categorized / deconflicted / pooled / dialectic), and set the maximum deconfliction rounds. Each field supplied is persisted and survives restarts/reloads, same as setup_council's tier choices; a field left unset is untouched.

Send a question to the model council and get a structured response. Mode: individual (each model answers separately), categorized (judge groups responses into agreement/complementary/conflicting), deconflicted (iterative loop — judge orchestrates re-questioning until conflicts resolve, returns a deconfliction score 0–100%), pooled (Delphi-style — members reconsider against a neutral, deduplicated, attribution-free pool of answers; no winner is forced, so genuine divergence is preserved), or dialectic (thesis/antithesis/synthesis — members defend their pick and critique the rest, the judge compiles a pros/cons dossier per option, then members re-select a ranked top-3). Attach images to ask a vision question — only auto-detected vision-capable members are queried; the rest are skipped and reported in visionRouting. For a repo review, pass git_ref (e.g. "uncommitted", "main..HEAD") instead of hand-listing files — the server runs git diff locally and attaches it as context. For a full repo-wide review (not just a diff), full_repo_access (default false, WARNING: read access to the whole repo — see its param description) grants claude-cli/codex-cli members read-only browse/read access.

Start a council run in the background and return a job_id immediately, so a long deconfliction/dialectic run (or a slow local model) does not block. Same inputs as ask_council (mode, context, files, etc.). Poll get_council_result with the job_id to fetch the answer when ready. Jobs are in-memory and do not survive a server reload.

Fetch a background council run started with ask_council_async. Pass job_id to get its status (running | done | error) and, when done, the full result. Omit job_id (or set list=true) to list recent jobs.

Return the current council configuration: member models, judge model, response mode, and max deconfliction rounds.

Report the detected environment and current setup: local Ollama models, whether Ollama cloud is reachable on this plan, whether Claude and Codex are logged in, whether Grok CLI is installed but fail-closed, the current council members, resolved subscription tiers, per-provider concurrency, and a quota warning. Use this as the welcome/status readout — it works in every client and install method.

Set subscription tiers, then re-detect and auto-populate the council with everything usable. Tiers gate cloud availability and per-provider concurrency: chatgpt (free|plus|pro5x|pro20x), claude (free|pro|max5x|max20x), grok (free|supergrok|premiumplus|heavy), ollama (free|pro|max). Choices persist across reloads. Note: registering a NEW subscription provider or changing concurrency takes full effect after a reload.

Set the per-completion wall-clock timeouts (ms) for council calls, persisted across reloads and overriding the REQUEST_TIMEOUT_MS / REPO_REQUEST_TIMEOUT_MS env defaults. run_timeout_ms applies to text-only ask_council calls; repo_timeout_ms applies when full_repo_access is set (repo-reading completions run longer). Omit either to leave it unchanged. Raise these when a member answer is cut mid-generation (the result then carries a timeoutNotice). Returns the now-effective values. A reload is NOT required — takes effect on the next ask_council.