Two-tier runtime model, full provider matrix, and a System health tab - #40
Merged
Conversation
…nt runtime The openai provider now runs on the official openai SDK; a new gemini provider speaks the native generateContent API via @google/genai (function calling, inline images, raw-JSON-Schema tool declarations). openai-compat/litellm keep the minimal wire client. Provider clients plug in below the ritsu-agent loop, so the approval gate still fronts every gated tool call. Sampling params are sent only when set in provider_options so reasoning models keep their server-side defaults.
…ovider matrix Agent definitions now declare an explicit runtime + provider pair. direct = a vendor agent runtime riding a subscription (claude today); api = ritsu's own tool loop against a metered model API. Existing databases migrate automatically (the legacy dispatcher column is rebuilt away; openai-compat rows become openrouter). api tier: anthropic (official SDK, Messages API), openai (official SDK), gemini (official SDK, native generateContent), xai/Grok (api.x.ai, xAI's documented OpenAI-compatible path), openrouter, litellm (key optional, falls back to connector credentials), and custom base_url endpoints. Keyless endpoints send no Authorization header. The admin form gains a runtime selector with per-tier provider lists, and the test pane can exercise api-runtime providers directly.
GET /admin/api/health probes the core runtime (DB, Claude session), every stored provider key (model-list endpoints, no token spend), and each configured connector (LiteLLM, Flashback, SearXNG, ingest), with configured/partial reporting for email/social. 4s timeouts, parallel, no secret material in results.
|
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What
runtime(direct= vendor agent runtime on a subscription,claudetoday;api= ritsu's own tool loop on a metered key) +provider. Replaces the olddispatcherfield and the provider/api_key_ref inference; existing databases migrate automatically (one-shot table rebuild, covered by tests).anthropic/openai/geminivia their official SDKs,xai(Grok via api.x.ai),openrouter,litellm(key optional — falls back to connector credentials),custombase_url. Provider clients plug in below the ritsu-agent loop, so the approval gate stays the enforcement point. Sampling params sent only when set inprovider_options.Notes
openai-compatmaps toopenrouteron migration.Testing