refactor: remove retired agent provider integration

This commit is contained in:
2026-11-12 15:53:33 -08:00
parent d88018fd64
commit 91c7c0310e
42 changed files with 271 additions and 245 deletions
+13 -12
View File
@@ -37,14 +37,13 @@ Chat upload limits:
"openai": { "models": ["chat-latest", "gpt-4.1-mini"], "loadedAt": "2026-02-14T00:00:00.000Z", "error": null },
"anthropic": { "models": ["claude-3-5-sonnet-latest"], "loadedAt": null, "error": null },
"xai": { "models": ["grok-3-mini"], "loadedAt": null, "error": null },
"gemini": { "models": ["gemini-3.5-flash"], "loadedAt": null, "error": null },
"hermes-agent": { "models": ["hermes-agent"], "loadedAt": null, "error": null }
"gemini": { "models": ["gemini-3.5-flash"], "loadedAt": null, "error": null }
}
}
```
- Selectable providers are `openai`, `anthropic`, `xai`, and `gemini`. Unsupported provider IDs are rejected with HTTP 400 by chat creation, chat completion, and Quick Question endpoints. Sybil Agent runs through its dedicated room or an inline `@sybil` mention; it is not a model-catalog provider.
- OpenAI model lists are filtered to models that are expected to work with the backend's Responses API implementation, including the exact `chat-latest` alias when returned by OpenAI. Audio, image-generation, embedding, moderation, and other specialized model ids remain excluded. See [OpenAI's Chat Latest documentation](https://developers.openai.com/api/docs/models/chat-latest) for supported capabilities.
- Gemini model lists are loaded from Google's native Models API and filtered to Gemini `generateContent` model ids.
- `hermes-agent` is included only when `HERMES_AGENT_API_KEY` is configured. Set it to Hermes `API_SERVER_KEY`, or any non-empty value if that local server does not require auth. `HERMES_AGENT_API_BASE_URL` defaults to `http://127.0.0.1:8642/v1`; set `HERMES_AGENT_MODEL` only when you need an additional fallback/override model id.
- The backend loads provider model lists at startup and refreshes them about once every 24 hours. If a later provider refresh fails, the response keeps the last loaded model list for that provider and sets `error` to the latest failure message.
## Chat Tools
@@ -148,7 +147,7 @@ History search (`q`) notes:
```json
{
"title": "optional title",
"provider": "optional openai|anthropic|xai|gemini|hermes-agent",
"provider": "optional openai|anthropic|xai|gemini",
"model": "optional model id",
"additionalSystemPrompt": "optional stored system prompt",
"enabledTools": ["web_search", "fetch_url"],
@@ -286,7 +285,7 @@ Notes:
{
"chatId": "optional-chat-id",
"clientRequestId": "optional-client-generated-id",
"provider": "openai|anthropic|xai|gemini|hermes-agent",
"provider": "openai|anthropic|xai|gemini",
"model": "string",
"messages": [
{
@@ -337,7 +336,8 @@ Behavior notes:
- If `chatId` is present, server validates chat existence.
- `clientRequestId` is an optional stable identifier for retries of a persisted inline Agent completion.
- For a persisted ordinary chat, a literal case-insensitive `@sybil` token in the latest user message invokes the existing persistent Sybil Agent instead of the selected provider for that turn. Email addresses, code spans/fences, indented code, and Markdown quote lines do not trigger it. The message and Agent reply remain visible in the chat; both are marked as Agent asides and are excluded by the backend from later selected-provider conversation history. The Agent root request is not projected into the visible Agent transcript.
- The completion response identifies an inline Agent reply with `provider: "hermes-agent"`, `model: "sybil-agent"`, and `message.name: "Sybil Agent"`, `message.agentAside: true`.
- The completion response identifies an inline Agent reply with `provider: "sybil-agent"`, `model: "sybil-agent"`, and `message.name: "Sybil Agent"`, `message.agentAside: true`.
- The response provider identity can therefore differ from the request's selectable provider. Clients must accept `sybil-agent` in completion responses without adding it to provider selectors.
- Agent asides are identified by backend-stored `metadata.agentAside`, not client-supplied metadata. The ordinary-chat list/history APIs above support Agent tools reading the complete visible thread, including prior asides and attachment data, with cursor pagination.
- Before the Agent is invoked, the backend paginates and supplies the complete safe visible-thread projection as required initial context; the Agent does not have to choose `read_chat_history` for the current thread. User/assistant asides and full supported attachment representations are included, while system/tool rows and unrelated metadata are excluded. Historical content remains untrusted context, not new authority.
- The initial context transfer is limited to 30 MiB. A stored text attachment marked truncated, incomplete history, or a transfer/provider context-limit failure returns an explicit error and no answer from partial history; providers may reject a complete thread that exceeds their model context window.
@@ -355,10 +355,9 @@ Behavior notes:
- For `anthropic`, backend calls Anthropic's Messages API and enables internal tool use with Anthropic `tool_use`/`tool_result` content blocks.
- For `xai`, backend calls xAI's OpenAI-compatible Chat Completions API and enables internal tool use with the same internal system instruction.
- For `gemini`, backend calls Google's native Gemini `generateContent` API and enables internal tool use with Gemini function calling.
- For `hermes-agent`, backend calls the configured Hermes Agent OpenAI-compatible Chat Completions API without adding Sybil-managed tool definitions; Hermes Agent handles its own tools server-side.
- For `openai`, image attachments are sent as Responses `input_image` items and text attachments are sent as `input_text` items.
- For `gemini`, image attachments are sent as native Gemini `inlineData` parts and text attachments are sent as text parts.
- For `xai` and `hermes-agent`, image attachments are sent as Chat Completions content parts alongside text.
- For `xai`, image attachments are sent as Chat Completions content parts alongside text.
- For `openai`, Responses calls that can enter the server-managed tool loop use `store: true` so reasoning and function-call items can be passed between tool rounds.
- For `anthropic`, image attachments are sent as Messages API `image` blocks using base64 source data; text attachments are added as `text` blocks.
- Available Sybil-managed tool calls for `openai`, `anthropic`, `xai`, and `gemini`: `web_search` and `fetch_url`. When `CHAT_CODEX_TOOL_ENABLED=true`, `codex_exec` is also available. When `CHAT_SHELL_TOOL_ENABLED=true`, `shell_exec` is also available.
@@ -481,6 +480,8 @@ Behavior notes:
## Type Shapes
Saved selections for unsupported provider IDs are returned as `null` for both the provider and its paired model. Chat messages and historical call records are preserved; an unsupported `calls[].provider` is returned as `null`, while the original call request/response snapshots remain unchanged. Clients should choose a supported provider before continuing those chats.
`ChatSummary`
```json
{
@@ -492,9 +493,9 @@ Behavior notes:
"updatedAt": "...",
"starred": false,
"starredAt": null,
"initiatedProvider": "openai|anthropic|xai|gemini|hermes-agent|null",
"initiatedProvider": "openai|anthropic|xai|gemini|null",
"initiatedModel": "string|null",
"lastUsedProvider": "openai|anthropic|xai|gemini|hermes-agent|null",
"lastUsedProvider": "openai|anthropic|xai|gemini|null",
"lastUsedModel": "string|null",
"additionalSystemPrompt": null,
"enabledTools": ["web_search", "fetch_url"]
@@ -546,9 +547,9 @@ Behavior notes:
"updatedAt": "...",
"starred": false,
"starredAt": null,
"initiatedProvider": "openai|anthropic|xai|gemini|hermes-agent|null",
"initiatedProvider": "openai|anthropic|xai|gemini|null",
"initiatedModel": "string|null",
"lastUsedProvider": "openai|anthropic|xai|gemini|hermes-agent|null",
"lastUsedProvider": "openai|anthropic|xai|gemini|null",
"lastUsedModel": "string|null",
"additionalSystemPrompt": null,
"enabledTools": ["web_search", "fetch_url"],
+5 -5
View File
@@ -26,7 +26,7 @@ Web presentation note: `/lite` uses these same chat streams and existing search
"chatId": "optional-chat-id",
"persist": true,
"clientRequestId": "optional-client-generated-id",
"provider": "openai|anthropic|xai|gemini|hermes-agent",
"provider": "openai|anthropic|xai|gemini",
"model": "string",
"messages": [
{
@@ -76,7 +76,8 @@ Notes:
- `maxTokens` is optional. For `anthropic`, when omitted the backend requests the selected model's maximum output token limit from Anthropic's Models API and uses that as `max_tokens`; if the model limit cannot be loaded, the fallback is 128000. For other providers, omitted `maxTokens` is not sent as an explicit cap.
- Attachments are optional and are persisted under `message.metadata.attachments` on stored user messages when `persist` is `true`.
- For a persisted ordinary chat, a literal case-insensitive `@sybil` token in the latest user message invokes the existing persistent Sybil Agent instead of the selected provider. Email addresses, code spans/fences, indented code, and Markdown quote lines do not trigger it. The visible user message and inline Agent reply remain in the ordinary chat and are marked with backend-owned `metadata.agentAside`; the backend excludes both from later selected-provider context without deleting the visible rows. This routing does not apply to the special Agent tab or Quick Questions.
- The `meta` event for an inline Agent reply contains `provider: "hermes-agent"`, `model: "sybil-agent"`, and `agentAside: true`. The selected ordinary provider is not run for that turn. Clients should render the response with the existing Sybil Agent identity and retain aside metadata on the optimistic streaming message until the canonical REST refresh.
- The `meta` event for an inline Agent reply contains `provider: "sybil-agent"`, `model: "sybil-agent"`, and `agentAside: true`. The selected ordinary provider is not run for that turn. Clients should render the response with the existing Sybil Agent identity and retain aside metadata on the optimistic streaming message until the canonical REST refresh.
- `sybil-agent` is a response identity, not an accepted request provider or a model-picker option. Decode completion metadata independently from the selectable provider list. Active-stream attachment and completed-request retries use the same identity.
- Before the Agent request starts, the backend loads every page of the safe visible thread and sends it as required initial context, including previous asides and supported attachment contents. This is not an optional `read_chat_history` tool call. Hidden system/tool rows and unrelated metadata are excluded; transcript data is untrusted context. If history/attachment projection is incomplete or exceeds the 30 MiB transfer limit, or the Agent model rejects the complete context, the stream reports an error and emits no partial-context answer.
Persisted chat streams with a `chatId` are backend-owned active runs:
@@ -105,7 +106,7 @@ This endpoint is intended for clients that restored an active `chatId` from `GET
Request body:
```json
{
"provider": "openai|anthropic|xai|gemini|hermes-agent",
"provider": "openai|anthropic|xai|gemini",
"model": "string",
"question": "What is the capital of France?",
"enabledTools": ["web_search", "fetch_url"],
@@ -218,10 +219,9 @@ Terminal tool-call event:
- `xai`: backend uses xAI's OpenAI-compatible Chat Completions API and may execute the same internal tool calls before producing final text.
- `gemini`: backend uses Google's native Gemini `streamGenerateContent` API and may execute the same internal tool calls before producing final text.
- `fetch_url` sends browser-like navigation headers for outbound URL requests to reduce false 403s from sites that reject generic server clients. It converts HTML to text and extracts selectable PDF text page by page (without OCR), including when the response media type is inaccurate but the PDF signature is present. URL downloads are limited to 10 MiB, PDF extraction to 30 pages per call and an isolated worker terminated after 8 seconds, and accumulated PDF text to `maxCharacters` (default 12,000; maximum 50,000), plus a brief truncation hint. PDF tool results include `pdfStatus`, `pageCount`, and selected `pages`; call with `pages:[...]` to retrieve specific pages. Encrypted, page-free, and text-free PDFs return explicit status and text with `ok:false`. Oversized responses fail with a size-limit error.
- `hermes-agent`: backend uses the configured Hermes Agent OpenAI-compatible Chat Completions API. Sybil does not add its own tool definitions for this provider; Hermes Agent handles its own tools server-side. Custom Hermes stream events are normalized away unless they produce text deltas in this SSE contract.
- `openai`: image attachments are sent as Responses `input_image` items; text attachments are sent as `input_text` items.
- `gemini`: image attachments are sent as native Gemini `inlineData` parts; text attachments are inlined as text parts.
- `xai` and `hermes-agent`: image attachments are sent as Chat Completions content parts; text attachments are inlined as text parts.
- `xai`: image attachments are sent as Chat Completions content parts; text attachments are inlined as text parts.
- `openai`: Responses calls that can enter the server-managed tool loop use `store: true` so reasoning and function-call items can be passed between tool rounds.
- `anthropic`: streamed via event stream; emits `delta` from `content_block_delta` with `text_delta`, and emits normalized `tool_call` SSE events when Anthropic `tool_use` blocks are executed. Image attachments are sent as base64 `image` blocks and text attachments are appended as `text` blocks.
- `web_search` uses `CHAT_WEB_SEARCH_ENGINE`: `exa` (default), `brave` (requires `BRAVE_SEARCH_API_KEY`), or `searxng` (requires `SEARXNG_BASE_URL`; the instance must allow `format=json`). This only affects chat-mode tool calls, not search-mode endpoints.