docs(graphiti): note beta status and OpenAI-only provider limitation

[mason5052]

Summary

Documents that the Graphiti knowledge graph integration is currently a beta feature, and that the bundled container today only consumes OpenAI-compatible LLM credentials. Adds the note in two places (README and backend/docs/config.md) without changing any runtime behavior.

Problem

Issue #187 raised that Graphiti's current provider story is not obvious to users. PentAGI itself is configured against many LLM providers (Anthropic, Google AI / Gemini, AWS Bedrock, DeepSeek, GLM, Kimi, Qwen, OpenAI, Ollama), but the bundled vxcontrol/graphiti container in docker-compose-graphiti.yml only reads OPENAI_API_KEY and a single OPEN_AI_SERVER_URL (default https://api.openai.com/v1) for entity extraction, with a fixed model name from GRAPHITI_MODEL_NAME (default gpt-5-mini).

Today neither the README's Graphiti section nor backend/docs/config.md's Graphiti settings call this out, and they do not flag the integration as beta. Users have to read docker-compose-graphiti.yml to understand the constraint.

Solution

Add a small, accurate, English-only note in two surfaces:

• README.md — Adds an [!IMPORTANT] beta callout at the top of the Knowledge Graph Integration (Graphiti) section and a new Current Limitations subsection covering:
  • OpenAI-compatible LLM only (with the relevant env vars and the list of unaffected PentAGI providers),
  • single fixed model per deployment,
  • independent billing from the main flow LLM,
  • no in-app graph explorer yet (Neo4j Browser / Swagger UI are the supported surfaces),
  • explicit fallback: leave GRAPHITI_ENABLED=false if the deployment cannot reach an OpenAI-compatible endpoint.
• backend/docs/config.md — Mirrors the same beta callout at the top of Graphiti Knowledge Graph Settings and adds a Current Limitations (Beta) subsection so config-focused readers see the same message.

The commit is documentation only; no runtime, schema, or env-var behavior changes.

User Impact

• Operators evaluating Graphiti can see the OpenAI-compatible-LLM constraint and the beta status before they enable it in production.
• Users running PentAGI on a non-OpenAI provider learn the documented fallback (GRAPHITI_ENABLED=false) and that the rest of the stack still works.
• The note is honest about scope: it does not promise multi-provider Graphiti work; it documents what ships today and points at what the user can choose.

Test Plan

• git diff --stat shows only README.md and backend/docs/config.md changed.
• Markdown renders without broken anchor, callout, or code-fence syntax.
• Constraints quoted in the note match docker-compose-graphiti.yml on feature/next-release (env vars OPENAI_API_KEY, OPENAI_BASE_URL from OPEN_AI_SERVER_URL, default https://api.openai.com/v1, MODEL_NAME from GRAPHITI_MODEL_NAME, default gpt-5-mini).
• No code or tests changed.

Refs #187

[Copilot]

Pull request overview

This PR updates PentAGI documentation to clearly flag the Graphiti knowledge graph integration as a beta feature and to document its current provider limitations (notably that the bundled Graphiti container uses OpenAI-compatible credentials/config only), without changing runtime behavior.

Changes:

• Added a beta/limitations callout to the Graphiti section in README.md.
• Added a “Current Limitations” subsection in README.md describing provider/model/UI/billing constraints.
• Mirrored the same beta/limitations messaging in backend/docs/config.md.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 2 comments.

File	Description
README.md	Adds Graphiti beta callout + a “Current Limitations” subsection for operators.
backend/docs/config.md	Adds the same beta/limitations note to the Graphiti configuration documentation.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.