production AI platform kernel: LLM abstraction, workflow engine, RAG, MCP server, FastAPI, evaluation framework, closed-loop feedback
View on GitHub ↗The platform kernel for an AI content-automation system: one versioned, tested Python package that four content pipelines depend on, plus the service layer, persistence, observability, and an MCP server that turn those pipelines into a production platform.
It began as a way to eliminate ~1,500 lines of logic copy-pasted across four repositories. It has since grown into the shared foundation for LLM access, retrieval, agent orchestration, workflow execution, cost tracking, and a REST API — while keeping every addition justified by a real need (see docs/adr/).
Four interfaces, one engine — the CLI pipelines, the REST API, the MCP server, and the web dashboard all drive the same core. No business logic is duplicated across them.
Module Responsibility ------ contentcore.llm.LLMProvider Provider-agnostic text generation (Claude/OpenAI/Gemini), lazy SDK imports, retry + backoff, normalized responses, telemetry hook contentcore.workflow Step-based workflow engine — named stages, per-stage retry, optional-stage skipping, run reports contentcore.retry Retry decorator; retries only listed exception types (auth errors fail fast) contentcore.rag Semantic perspective retrieval — embeddings + cosine search, fingerprint-cached index, graceful fallback contentcore.agents Multi-agent story production — Director → Writer/Visual/Audio with a bounded single-revision critique loop contentcore.telemetry Per-call LLM token/cost/latency recording, structured JSON logging, optional Langfuse export contentcore.db SQLAlchemy persistence — workflow runs, step records, LLM usage, generated content c
From the project README.
Add the radar badge to your README — it shows your project was picked up by MCP Radar and links to this page:
[](https://mcp.liqiwa.com/s/sai-chaithanya-navuluri--content-core.html)
Self-hosted documentation RAG pipeline that serves semantic search to LLM agents over the Model Context Protocol.
SujalXplores/Agent-KCode-enforced incident-response agent that investigates SigNoz alerts, publishes evidence-backed root-cause claims, and gates every action behind a safety policy. Built for Agents of SigNoz hackathon.
graphersal/graphersalIn-memory graph database and Gremlin traversal engine in Rust: browser playground, CLI, MCP server for AI agents, durable store, Python bindings.
2akouwu/reverifyStop your AI from making things up — it proposes, deterministic tools decide, every claim checked against ground truth with evidence. Grounded facts and context survive resets. Reverse engineering is
bybit-exchange/kaasTurn scattered notes, docs and transcripts into a queryable Markdown wiki — an LLM knowledge-base compiler with MCP access, no embeddings, self-hosted.
RayenMlayeh/jevcrawlMCP server and CLI that crawls websites for AI agents: finds the answer page 2x more often than Crawl4AI, 10x cheaper per answer than letting Claude crawl.
The top new MCP servers of the week, every Monday. No spam, unsubscribe anytime.