Join a video meeting as an AI bot with voice, avatar, and screenshare across four operating modes.
by johnpatternai21 installs8 stars
Design & media
sith-choir
Try itCross-model intellect relay across a mesh of stored API keys (multi-provider, multi-key). The conductor farms work out to a choir of models reachable via EVERY stored credential — each provider's api_key and each alt_api_keys entry is its own health-tracked voice with automatic per-key failover — and models that need help are answered using other providers' keys (e.g. a Mistral-key model that is rate-limited or unsure consults Gemini and llm7.io keys). Core moves: (1) weak/cheap models sing candidates in parallel and a reasoning model aggregates them (Mixture-of-Agents, arXiv:2406.04692); (2) vision relay — text-only callers hand images to vision-capable voices (arXiv:2502.16428); (3) mandatory escalation — low confidence, errors or rate limits route to a reasoning court from other providers (FrugalGPT cascade, arXiv:2305.05176); (4) judge panels score 0-10 and a persistent reward ledger punishes weak voices while key-health cooldowns rotate load (arXiv:2306.05685). Plus self-consisten
What it does
🦞 sith-choir — Cross-Model Intellect Relay
The skill document
🦞 sith-choir — Cross-Model Intellect Relay
Let the choir sing; let the Sith listen. This skill turns one conductor agent into a capitalist overlord of a choir of other models: it farms questions out to the cheapest/weakest models, drains the best intelligence from the strongest reasoners, relays vision work to models that can actually see, judges every answer with a panel, and keeps a reward ledger that punishes weak singers and promotes strong ones.
Not a toy, not a metaphor for anything nefarious: every move maps to a
peer-reviewed or industry-validated multi-agent technique (see
references/design.md).
When to use
- You have many provider API keys (e.g. a 55-key workspace) and want the best answer instead of one lucky model.
- A text-only model is handed an image — it cannot see, so it must ask a
vision-capable model to do the job (
vision). - A model is error-prone or low-confidence — policy says it MUST escalate to
the best reasoning models (
escalate). - You want answers judged by other models and good/bad performers tracked
across sessions (
judge+harvest+ledger).
Quick start
# In an arena.ai workspace with the repo's secrets store:
python3 scripts/choir.py list --use-workspace
python3 scripts/choir.py moa "explain the CAP theorem" --use-workspace
python3 scripts/choir.py vision --image screenshot.png \
--question "what is on this screen?" --use-workspace --synthesize
python3 scripts/choir.py harvest "solve this math problem" --use-workspace --json
python3 scripts/choir.py ledger --use-workspace
Anywhere else, point at your own credentials file:
python3 scripts/choir.py list --credentials ~/.config/sith-choir/credentials.json
Credentials JSON: {"providers":{"gemini":{"api_key":"..."}, ...}}. Optional
per provider: base, kind (openai|gemini), model, tags.
Commands
| Command | What it does | Based on |
|---|---|---|
list | Show roster + reward ledger | — |
ask "Q" [--from prov/model] | Single direct answer (cheap lane) | — |
moa "Q" [--proposers N] | Parallel weak/cheap proposals, strong aggregator merges | MoA 2406.04692 |
vision --image F --question Q [--synthesize] | Text-only caller relays the image to vision models; optional reasoning synthesis pass | 2502.16428 / 2502.13923 |
escalate "Q" [--force] [--require-conf] | Primary answers with a CONF: line; below 0.7 (or forced) → mandatory reasoning court | FrugalGPT cascade 2305.05176 |
judge --question Q --answer @file [--members ...] | N judges score 0-10; reward ≥7.5, punish ≤4 | LLM-as-Judge 2306.05685 |
harvest "Q" | propose → aggregate → judge → ledger in one move | MoA + judge |
ledger [--reset] | Persistent reward/punishment history per member | — |
selftest | Offline test suite (sandboxed, no network, no state) | — |
Every command accepts --json. Providers are tried with one 429 backoff, then
clearly reported — no silent empty answers, ever.
Capability routing
classify() tags each model with vision / reasoning / fast / longctx
by id heuristics. It is a hint: override per provider in your credentials
file with "tags": ["vision"]. vision uses members tagged vision;
moa/escalate/harvest prefer reasoning members for aggregation/court.
Related skills
Stores durable facts in a categorized, plain-markdown vault on disk, alongside your agent's built-in memory.
by Iván555 installs18 stars
Trade crypto, manage a multi-chain wallet, and query an AI analyst from one CLI.
by lowesyang162 installs109 stars
Find why your productivity system keeps failing, then apply the smallest fix — capacity math, bottleneck routing, durable local notes.
by Iván854 installs69 stars
Generate and edit Draw.io, Mermaid, and Excalidraw diagrams from natural language using a structured JSON spec.
by nssa.io1.0k installs47 stars
Query and manage Linear issues, projects, teams, cycles, labels, and comments through a managed OAuth GraphQL endpoint.
by byungkyu518 installs18 stars
More from orionshaowswmw
Browse all skillsDetect and repair partially wiped agent workspaces with integrity checks, signed manifests, guarded restore recipes, bounded local recovery state, and explicit off-box sync. Use when files, scripts, trees, models, or build outputs disappear or lose integrity between turns.
by orionshaowswmw1 installs1 stars
Seven offline mechanisms against slow/stale/zombie/sycophantic agent turns: prompt compaction, request fencing, zombie detection, CAPTCHA triage, anti-sycophancy spine, delivery register, invention quarry. Use when chat feels laggy, reconnects surface old answers, long chats degrade, or the agent caves under contradiction. JSON contracts; state per-agent under ~/.arena_turn; no network, no sudo.
by orionshaowswmw2 installs
Quota-aware LLM router that squeezes maximum usable AI out of free-tier API keys across Gemini, Mistral, OpenRouter, Kilo and Cerebras plus any OpenAI-compatible endpoint (including local Ollama/llama.cpp/vLLM). Probes every model on every key, measures real quality and real published rate limits, then routes each request to the cheapest model that can do the job — spending abundant capacity first and reserving scarce daily quota for when it is actually needed. Persists cooldowns to disk so a 429 discovered in one process is respected by the next. Use when an agent must make many LLM calls on free keys without hitting rate limits, when "all models failed", or when deciding which of several provider keys to use for a task.
by orionshaowswmw2 installs
Opt-in, model-neutral guidance for evidence-aware, dignified AI communication, with a compact response contract and offline deterministic text audit. It never injects prompts, edits host configuration, calls networks, reads secrets, or treats heuristics as truth.
by orionshaowswmw2 installs
Iran Chemical Database — live, dated, auditable, BEST-EFFORT index of chemical offerings in configured public Iranian supplier catalogues (websites + public Telegram channels). HTTrack/WooCommerce-REST/Telegram mirroring → local-only parsing → RDKit/PubChem/CAS-validated PostgreSQL with FastAPI + Streamlit. Fail-closed Iranian-suppliers-only country gate; coverage measured and published, never claimed complete. Installation = software + queued crawl, not a populated dataset. Ships a 1399-molecule CID-unique confirmed-organic seed baseline (v2.22, 2026-08-27: v2.19 primary + live Telegram/WooCommerce/sitemap crawl + 5-model fleet normalization, every new identity PubChem-confirmed). For academic procurement research.
by orionshaowswmw3 installs
Model-agnostic, agent-agnostic fidelity-first pipeline converting operator-authorized Persian/English RTL lecture PDFs into offline HTML study guides — recall-first dual OCR (PyMuPDF + Tesseract fas+eng PSM ensemble), rendered-page evidence, multi-model correction, session-grounded enrichment (tables/flashcards/quizzes/mnemonics/summaries/scenarios), measured fidelity, QA gates, verified ZIP. v1.5.0 runs on ANY model family through 8 API dialects (OpenAI, Responses, Gemini, Anthropic, Cohere, Ollama, HuggingFace, offline mock) or with no model at all, auto-discovers providers from the host agent's environment, self-heals provider quirks and model retirements, and exposes one deterministic CLI/MCP entrypoint plus cross-model consensus so different agents reproduce the same intended result.
by orionshaowswmw2 installs