Sol-37 / Public API

Privacy Policy

Sol exposes read-only semantic search, grounded chat, and live status retrieval over public endpoints at sol.system42.one. This policy covers both the legacy GPT Action surface and the newer Sol MCP app surface.

What is sent

• The text of your search query or chat query.
• Optional page-context fields supplied by the caller, such as page title, headings, or page content summaries.
• Optional attached image payloads for multimodal chat turns when the caller uses the vision-capable chat route.
• Current site metrics, sensor readouts, and local service telemetry may also be read on the server and folded into text-only responses as ambient runtime context.
• Basic request metadata required to serve the API or MCP connector, such as timestamps and standard web-server logs.

What the API does

• Looks up semantically similar text chunks in the existing Sol knowledge index.
• Returns matching excerpts and metadata such as path, chunk, and score.
• May synthesize a grounded chat response using local model infrastructure plus retrieved knowledge or supplied page context.
• May also synthesize text-only responses using current runtime context from site metrics and local service telemetry when available.
• May describe attached images using the local vision backend when the chat request includes image data.
• Does not mutate the knowledge base, trigger automations, or perform user actions.
• The MCP app surface exposes read-only tools only: knowledge search, document fetch, grounded chat, and live status reads.

Storage and processing

• Queries may appear in local server logs used for operations and debugging.
• Attached image payloads are processed to answer the current request and may appear transiently in in-memory backend request handling and debug logs.
• Ambient runtime context is sourced locally from the Sol host, including current sensor readouts and selected service status/console telemetry exposed by the local dashboard stack.
• The knowledge index is hosted on the server running the Sol site.
• Depending on the configured embedding backend, query text may be sent to an external embedding provider to compute the retrieval query vector.
• Grounded chat responses are generated by local model services running on the Sol host.

Current backend note

The live service currently uses the same embedding pipeline used by Sol and MasterBot. If that pipeline is configured to use OpenAI embeddings, your query text is sent to OpenAI for query embedding generation before local similarity search is performed.

Contact

• Name: David Lones
• Email: [email protected]
• Public guestbook: https://sol.system42.one/gui?i=programs%2Flogbook.html