lywsvip/ironclaw
1
0
Fork 0
mirror of https://github.com/nearai/ironclaw.git synced 2026-09-03 08:06:01 +08:00
Code Issues Packages Projects Releases Wiki Activity
Files
11aa7e3f246bb993de0d50b38413afd883c301ed
ironclaw/providers.json
Illia Polosukhin 481a34a18e fix(llm): route DeepSeek, Gemini, and OpenRouter through dedicated rig-core clients (#3201, #3225) (#3326) 
* fix(llm): route DeepSeek and Gemini through dedicated rig-core clients (#3201, #3225)

DeepSeek thinking-mode (#3201) and Gemini API-key tool calling (#3225)
both failed deterministically on the second LLM turn:

- DeepSeek: HTTP 400 "The reasoning_content in the thinking mode must
  be passed back to the API"
- Gemini: HTTP 400 "Function call is missing a thought_signature in
  functionCall parts"

Both providers were configured in `providers.json` as
`protocol: "open_ai_completions"`, which dispatched them through
rig-core's generic OpenAI client. That client silently strips
`reasoning_content` from assistant messages and `thought_signature`
from tool calls when serializing the next turn — so the field that
the upstream API requires to be echoed back was never sent.

rig-core actually has dedicated `deepseek::Client` and `gemini::Client`
implementations that handle the round-trip correctly:

- `deepseek.rs:500-527` — captures `reasoning_content` from the response
  and writes it back onto the last assistant message in the next request.
- `gemini/completion.rs:1016` — round-trips `thought_signature` on every
  `ToolCall` via `ToolCall.signature`.

Add `ProviderProtocol::DeepSeek` and `ProviderProtocol::Gemini` variants
plus matching factory branches in `create_registry_provider`, and
update the two registry entries to use them. Both factories follow the
same shape as `create_anthropic_from_registry` and wrap the rig-core
client in our existing `RigAdapter` — no new HTTP code, no agent-loop
changes.

The `default_base_url` for both entries is now empty so the dedicated
clients use their built-in endpoints (`api.deepseek.com` /
`generativelanguage.googleapis.com`) rather than the OpenAI-compat
shims.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(llm): also route OpenRouter through dedicated rig-core client

Same bug class as DeepSeek (#3201) and Gemini (#3225): OpenRouter was
configured as `protocol: "open_ai_completions"`, dispatching it through
rig-core's generic OpenAI client. That client strips OpenRouter's
`reasoning`, `reasoning_details` (Summary / Encrypted / Text), and
per-tool-call signatures when serializing the next turn — breaking
tool calling for every thinking-mode model OpenRouter exposes (Claude
with thinking, OpenAI o-series, DeepSeek-R1, Gemini 2.5+, Qwen QwQ,
…).

rig-core's dedicated OpenRouter client (`openrouter/completion.rs`)
round-trips all of it correctly, including the `reasoning_details` ->
`tool_call.signature` mapping at lines 290-308.

Add `ProviderProtocol::OpenRouter`, a `create_openrouter_from_registry`
factory that preserves OpenRouter attribution headers (`HTTP-Referer`,
`X-Title`), and update the registry entry. Extend the regression test
to lock the routing for all three providers.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* ci: trigger refs/pull/3326/head regeneration

* fix(llm): round-trip reasoning + signatures through RigAdapter

Addresses review feedback on PR #3326. The original PR routed DeepSeek,
Gemini, and OpenRouter through rig-core's dedicated clients, but the
fix only worked inside rig-core. IronClaw's RigAdapter sits between
the agent loop and rig-core and was dropping AssistantContent::Reasoning
(DeepSeek `reasoning_content`) and per-tool-call `signature` (Gemini
`thought_signature`) on the response → IronClaw conversion. On the
next request it rebuilt rig messages without either field, so the
provider rejected the follow-up turn — same HTTP 400s the PR claimed
to fix.

Round-trip both fields end-to-end:
- Add `ChatMessage::reasoning` and `ToolCall.signature` (skip-serialized
  when None) plus `ToolCompletionResponse.reasoning` to carry artifacts
  out of the provider.
- Update `RigAdapter::extract_response` to capture both, and
  `convert_messages` to push `AssistantContent::Reasoning` and
  `ToolCall.signature` back when rebuilding rig messages on the next turn.
- Plumb response.reasoning through dispatcher, job worker, container
  worker, routine engine, and the orchestrator-worker proxy via a new
  `ChatMessage::with_reasoning` builder.

Other review fixes:
- Set `can_list_models: false` for Gemini in providers.json — setup
  wizard and web list-models handler don't speak native Gemini, so
  exposing the button reliably falls back to manual entry.
- Add `extra_headers_env: OPENROUTER_EXTRA_HEADERS` so users of the
  built-in `openrouter` backend can configure HTTP-Referer / X-Title
  attribution headers.
- New `sanitize_gemini_base_url` discards the legacy
  `…/v1beta/openai` shim URL persisted by pre-3225 installs in
  `llm_builtin_overrides[gemini].base_url`. Without this, upgraded
  users would hit `…/v1beta/openai/v1beta/models/...:generateContent`.
- Include `provider_id` in extra-header warning logs (OpenRouter +
  OpenAI-compat).

Regression tests:
- `reasoning_and_signature_round_trip_through_chat_message` simulates
  the 2-turn tool loop end-to-end.
- `chat_message_with_reasoning_drops_empty_input` locks the empty-input
  contract so we don't echo `reasoning_content: ""`.
- `sanitize_gemini_base_url_*` cover legacy-shim discard, empty input,
  and custom-endpoint pass-through.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-05-07 12:29:04 -07:00

451 lines
13 KiB
JSON
Raw Blame History

[
{
"id": "openai",
"aliases": [
"open_ai"
],
"protocol": "open_ai_completions",
"api_key_env": "OPENAI_API_KEY",
"api_key_required": true,
"base_url_env": "OPENAI_BASE_URL",
"model_env": "OPENAI_MODEL",
"default_model": "gpt-5-mini",
"description": "OpenAI GPT models (direct API)",
"unsupported_params": ["temperature"],
"setup": {
"kind": "api_key",
"secret_name": "llm_openai_api_key",
"key_url": "https://platform.openai.com/api-keys",
"display_name": "OpenAI",
"can_list_models": true
}
},
{
"id": "anthropic",
"aliases": [
"claude"
],
"protocol": "anthropic",
"api_key_env": "ANTHROPIC_API_KEY",
"api_key_required": true,
"base_url_env": "ANTHROPIC_BASE_URL",
"model_env": "ANTHROPIC_MODEL",
"default_model": "claude-sonnet-4-20250514",
"description": "Anthropic Claude models (direct API)",
"setup": {
"kind": "api_key",
"secret_name": "llm_anthropic_api_key",
"key_url": "https://console.anthropic.com/settings/keys",
"display_name": "Anthropic",
"can_list_models": true
}
},
{
"id": "ollama",
"aliases": [],
"protocol": "ollama",
"default_base_url": "http://localhost:11434",
"base_url_env": "OLLAMA_BASE_URL",
"model_env": "OLLAMA_MODEL",
"default_model": "llama3",
"description": "Local Ollama instance (no API key needed)",
"setup": {
"kind": "ollama",
"display_name": "Ollama",
"can_list_models": true
}
},
{
"id": "openai_compatible",
"aliases": [
"openai-compatible",
"compatible"
],
"protocol": "open_ai_completions",
"base_url_env": "LLM_BASE_URL",
"base_url_required": true,
"api_key_env": "LLM_API_KEY",
"api_key_required": false,
"model_env": "LLM_MODEL",
"default_model": "default",
"extra_headers_env": "LLM_EXTRA_HEADERS",
"description": "Custom OpenAI-compatible endpoint (vLLM, LiteLLM, etc.)",
"setup": {
"kind": "open_ai_compatible",
"secret_name": "llm_compatible_api_key",
"display_name": "OpenAI-compatible",
"can_list_models": false
}
},
{
"id": "github_copilot",
"aliases": [
"github-copilot",
"githubcopilot",
"copilot"
],
"protocol": "github_copilot",
"default_base_url": "https://api.githubcopilot.com",
"api_key_env": "GITHUB_COPILOT_TOKEN",
"api_key_required": true,
"model_env": "GITHUB_COPILOT_MODEL",
"default_model": "gpt-4o",
"extra_headers_env": "GITHUB_COPILOT_EXTRA_HEADERS",
"description": "GitHub Copilot Chat API (OAuth token from IDE sign-in)",
"setup": {
"kind": "api_key",
"secret_name": "llm_github_copilot_token",
"key_url": "https://docs.github.com/en/copilot",
"display_name": "GitHub Copilot",
"can_list_models": false
}
},
{
"id": "tinfoil",
"aliases": [],
"protocol": "open_ai_completions",
"default_base_url": "https://inference.tinfoil.sh/v1",
"api_key_env": "TINFOIL_API_KEY",
"api_key_required": true,
"model_env": "TINFOIL_MODEL",
"default_model": "kimi-k2-5",
"description": "Tinfoil private inference (hardware-attested TEE)",
"unsupported_params": ["temperature"],
"setup": {
"kind": "api_key",
"secret_name": "llm_tinfoil_api_key",
"key_url": "https://tinfoil.sh",
"display_name": "Tinfoil",
"can_list_models": false
}
},
{
"id": "openrouter",
"aliases": [
"open_router"
],
"protocol": "open_router",
"default_base_url": "",
"api_key_env": "OPENROUTER_API_KEY",
"api_key_required": true,
"model_env": "OPENROUTER_MODEL",
"extra_headers_env": "OPENROUTER_EXTRA_HEADERS",
"default_model": "openai/gpt-4o",
"description": "OpenRouter multi-provider gateway (200+ models, preserves reasoning across turns)",
"setup": {
"kind": "api_key",
"secret_name": "llm_openrouter_api_key",
"key_url": "https://openrouter.ai/settings/keys",
"display_name": "OpenRouter",
"can_list_models": false
}
},
{
"id": "groq",
"aliases": [],
"protocol": "open_ai_completions",
"default_base_url": "https://api.groq.com/openai/v1",
"api_key_env": "GROQ_API_KEY",
"api_key_required": true,
"model_env": "GROQ_MODEL",
"default_model": "llama-3.3-70b-versatile",
"description": "Groq LPU inference (ultra-fast)",
"setup": {
"kind": "api_key",
"secret_name": "llm_groq_api_key",
"key_url": "https://console.groq.com/keys",
"display_name": "Groq",
"can_list_models": true,
"models_filter": "chat"
}
},
{
"id": "nvidia",
"aliases": [
"nvidia_nim",
"nim"
],
"protocol": "open_ai_completions",
"default_base_url": "https://integrate.api.nvidia.com/v1",
"api_key_env": "NVIDIA_API_KEY",
"api_key_required": true,
"model_env": "NVIDIA_MODEL",
"default_model": "meta/llama-3.3-70b-instruct",
"description": "NVIDIA NIM API (high-performance inference)",
"setup": {
"kind": "api_key",
"secret_name": "llm_nvidia_api_key",
"key_url": "https://build.nvidia.com",
"display_name": "NVIDIA NIM",
"can_list_models": true
}
},
{
"id": "venice",
"aliases": [
"venice_ai",
"veniceai"
],
"protocol": "open_ai_completions",
"default_base_url": "https://api.venice.ai/api/v1",
"api_key_env": "VENICE_API_KEY",
"api_key_required": true,
"model_env": "VENICE_MODEL",
"default_model": "llama-3.3-70b",
"description": "Venice.ai privacy-focused inference",
"setup": {
"kind": "api_key",
"secret_name": "llm_venice_api_key",
"key_url": "https://venice.ai/settings/api",
"display_name": "Venice.ai",
"can_list_models": false
}
},
{
"id": "together",
"aliases": [
"together_ai",
"togetherai"
],
"protocol": "open_ai_completions",
"default_base_url": "https://api.together.xyz/v1",
"api_key_env": "TOGETHER_API_KEY",
"api_key_required": true,
"model_env": "TOGETHER_MODEL",
"default_model": "meta-llama/Llama-3-70b-chat-hf",
"description": "Together AI inference",
"setup": {
"kind": "api_key",
"secret_name": "llm_together_api_key",
"key_url": "https://api.together.ai/settings/api-keys",
"display_name": "Together AI",
"can_list_models": false
}
},
{
"id": "fireworks",
"aliases": [
"fireworks_ai"
],
"protocol": "open_ai_completions",
"default_base_url": "https://api.fireworks.ai/inference/v1",
"api_key_env": "FIREWORKS_API_KEY",
"api_key_required": true,
"model_env": "FIREWORKS_MODEL",
"default_model": "accounts/fireworks/models/llama-v3p1-70b-instruct",
"description": "Fireworks AI inference",
"setup": {
"kind": "api_key",
"secret_name": "llm_fireworks_api_key",
"key_url": "https://fireworks.ai/api-keys",
"display_name": "Fireworks AI",
"can_list_models": false
}
},
{
"id": "deepseek",
"aliases": [
"deep_seek"
],
"protocol": "deep_seek",
"default_base_url": "",
"api_key_env": "DEEPSEEK_API_KEY",
"api_key_required": true,
"model_env": "DEEPSEEK_MODEL",
"default_model": "deepseek-chat",
"description": "DeepSeek inference API (preserves reasoning_content for thinking-mode models)",
"setup": {
"kind": "api_key",
"secret_name": "llm_deepseek_api_key",
"key_url": "https://platform.deepseek.com/api_keys",
"display_name": "DeepSeek",
"can_list_models": false
}
},
{
"id": "zai",
"aliases": [
"bigmodel"
],
"protocol": "open_ai_completions",
"default_base_url": "https://api.z.ai/api/paas/v4",
"api_key_env": "ZAI_API_KEY",
"api_key_required": true,
"model_env": "ZAI_MODEL",
"default_model": "glm-5",
"description": "Z.AI GLM inference API",
"setup": {
"kind": "api_key",
"secret_name": "llm_zai_api_key",
"key_url": "https://z.ai/manage-apikey/apikey-list",
"display_name": "Z.AI",
"can_list_models": false
}
},
{
"id": "cerebras",
"aliases": [],
"protocol": "open_ai_completions",
"default_base_url": "https://api.cerebras.ai/v1",
"api_key_env": "CEREBRAS_API_KEY",
"api_key_required": true,
"model_env": "CEREBRAS_MODEL",
"default_model": "llama-3.3-70b",
"description": "Cerebras wafer-scale inference",
"setup": {
"kind": "api_key",
"secret_name": "llm_cerebras_api_key",
"key_url": "https://cloud.cerebras.ai",
"display_name": "Cerebras",
"can_list_models": false
}
},
{
"id": "sambanova",
"aliases": [
"samba_nova"
],
"protocol": "open_ai_completions",
"default_base_url": "https://api.sambanova.ai/v1",
"api_key_env": "SAMBANOVA_API_KEY",
"api_key_required": true,
"model_env": "SAMBANOVA_MODEL",
"default_model": "Meta-Llama-3.1-70B-Instruct",
"description": "SambaNova Cloud inference",
"setup": {
"kind": "api_key",
"secret_name": "llm_sambanova_api_key",
"key_url": "https://cloud.sambanova.ai/apis",
"display_name": "SambaNova",
"can_list_models": false
}
},
{
"id": "gemini",
"aliases": [
"google_gemini",
"google"
],
"protocol": "gemini",
"default_base_url": "",
"api_key_env": "GEMINI_API_KEY",
"api_key_required": true,
"model_env": "GEMINI_MODEL",
"default_model": "gemini-2.5-flash",
"description": "Google Gemini native API (preserves thought_signature on tool calls)",
"setup": {
"kind": "api_key",
"secret_name": "llm_gemini_api_key",
"key_url": "https://aistudio.google.com/app/apikey",
"display_name": "Google Gemini",
"can_list_models": false
}
},
{
"id": "ionet",
"aliases": [
"io_net",
"io.net"
],
"protocol": "open_ai_completions",
"default_base_url": "https://api.intelligence.io.solutions/api/v1",
"api_key_env": "IONET_API_KEY",
"api_key_required": true,
"model_env": "IONET_MODEL",
"default_model": "deepseek-coder-v2-instruct",
"description": "io.net Intelligence API",
"setup": {
"kind": "api_key",
"secret_name": "llm_ionet_api_key",
"key_url": "https://cloud.io.net/intelligence",
"display_name": "io.net",
"can_list_models": true
}
},
{
"id": "mistral",
"aliases": [
"mistral_ai",
"mistralai"
],
"protocol": "open_ai_completions",
"default_base_url": "https://api.mistral.ai/v1",
"api_key_env": "MISTRAL_API_KEY",
"api_key_required": true,
"model_env": "MISTRAL_MODEL",
"default_model": "mistral-large-latest",
"description": "Mistral AI API",
"setup": {
"kind": "api_key",
"secret_name": "llm_mistral_api_key",
"key_url": "https://console.mistral.ai/api-keys",
"display_name": "Mistral",
"can_list_models": true
}
},
{
"id": "yandex",
"aliases": [
"yandex_ai_studio",
"yandexgpt",
"yandex_gpt"
],
"protocol": "open_ai_completions",
"default_base_url": "https://ai.api.cloud.yandex.net/v1",
"api_key_env": "YANDEX_API_KEY",
"api_key_required": true,
"model_env": "YANDEX_MODEL",
"extra_headers_env": "YANDEX_EXTRA_HEADERS",
"default_model": "yandexgpt-lite",
"description": "Yandex AI Studio (YandexGPT)",
"setup": {
"kind": "api_key",
"secret_name": "llm_yandex_api_key",
"key_url": "https://aistudio.yandex.ru/platform/folders/",
"display_name": "Yandex AI Studio",
"can_list_models": true
}
},
{
"id": "minimax",
"aliases": [
"mini_max"
],
"protocol": "open_ai_completions",
"default_base_url": "https://api.minimax.io/v1",
"api_key_env": "MINIMAX_API_KEY",
"api_key_required": true,
"base_url_env": "MINIMAX_BASE_URL",
"model_env": "MINIMAX_MODEL",
"default_model": "MiniMax-M2.7",
"description": "MiniMax API (MiniMax-M2.7, MiniMax-M2.7-highspeed, MiniMax-M2.5 and MiniMax-M2.5-highspeed models)",
"setup": {
"kind": "api_key",
"secret_name": "llm_minimax_api_key",
"key_url": "https://platform.minimax.io",
"display_name": "MiniMax",
"can_list_models": false
}
},
{
"id": "cloudflare",
"aliases": [
"cloudflare_ai",
"cf_ai"
],
"protocol": "open_ai_completions",
"api_key_env": "CLOUDFLARE_API_KEY",
"api_key_required": true,
"base_url_env": "CLOUDFLARE_BASE_URL",
"model_env": "CLOUDFLARE_MODEL",
"default_model": "@cf/meta/llama-3.3-70b-instruct-fp8-fast",
"description": "Cloudflare Workers AI",
"setup": {
"kind": "open_ai_compatible",
"secret_name": "llm_cloudflare_api_key",
"display_name": "Cloudflare Workers AI",
"can_list_models": false
}
}
]
Reference in New Issue View Git Blame Copy Permalink