Conversational onboarding assistant: RAG over documents and code plus text-to-SQL behind an intent router. FastAPI + Next.js, Mistral-native, with a nightly eval loop and RLS-scoped answers.
Multi-agent RAG system built with LangGraph. Routes each message to chat, cited document Q&A, or a restricted Python agent that writes, runs and fixes code to calculate and chart numbers. FastAPI + Chainlit, session memory, Docker Compose, automated tests.
Reachy Mini harness: understand the sentence before you wake the model. Intent fast path, tool planner, measured latency
Enterprise-grade MEV-protected transaction router & fee sponsor for AI Agents on Solana
Cursor-Auto / Claude-tier-style serving for local GGUF models on Mac (M4 Max, 64 GB). FastAPI router fronts llama-swap + llama.cpp, classifying each request into a coder, planner, or uncensored-planner tier. OpenAI-compatible API, opencode integration, per-project subshell, one `llmstack` console-script.
ARIA · symbolic, learning AI (v1.3.1): 14 declarative skills + proof router + disciplined memory + self-audit + branded console/GUI. 100% local, no LLM, no cloud. / IA symbolique apprenante : 14 compétences déclaratives, routeur à preuve, interfaces de marque.
An intelligent multi-prompt routing and classification engine for SaaS customer support and workflow automation. Classifies incoming requests into discrete intent categories, evaluates urgency and sentiment, and dispatches queries to hyper-specialized expert prompts.
Android share-sheet router that normalizes shared text, URLs, and content URIs into JSON.
Zero-shot intent classification and out-of-scope detection. 150M model, 94.9% routing accuracy. ONNX int8 on CPU (160 ms p95) or MLX on Apple Silicon (~15-25 ms). Distilled from Laya into Ettin-150M.
中文意图路由器：词法打分+双门拒答+校准/留出集验收，附MCP server。零第三方依赖。
ARCHIVED — Merged into kochj23/nova. See https://github.com/kochj23/nova
To associate your repository with the intent-router topic, visit your repo's landing page and select "manage topics."