fix(providers): remove ~anthropic, skip ~ prefixes in sync script (#69)

* fix(providers): remove ~anthropic, skip ~ prefixes in sync script OpenRouter uses ~ prefixes for internal auto-routing aliases (e.g. ~anthropic). These are not real providers — they already route through openrouter.toml. The generated ~anthropic.toml was confusing (looked like a stale backup) and redundant with the existing openrouter provider. - Delete providers/~anthropic.toml - Skip provider IDs starting with ~ in sync-pricing.py --create-missing * fix(providers): remove morph, aider, kwaipilot - morph: specialized code-editing/patching tool, not a general LLM provider - aider: CLI meta-tool wrapper (base_url empty), redundant with claude-code/codex-cli/gemini-cli/qwen-code - kwaipilot: Kwai internal coding assistant routed via OpenRouter, niche * fix(sync): add morph/aider/kwaipilot to SKIP_PROVIDERS to prevent re-creation * feat(sync): merge OpenRouter-only providers into openrouter.toml Instead of generating standalone .toml files that just wrap the OpenRouter endpoint, merge their models directly into openrouter.toml with the standard 'openrouter/{provider}/{model}' ID convention. - Add _build_model_fields() and _model_lines() helpers to deduplicate model rendering between standalone and merged paths - Add merge_into_openrouter() that appends new models idempotently - generate_provider_toml() now only runs for providers in PROVIDER_API - --create-missing routes OpenRouter-only providers to merge_into_openrouter * fix(providers): remove 14 OpenRouter-only standalone files These providers have no direct public API and all route through openrouter.ai/api/v1. Per the new sync-pricing.py policy, their models will be merged into openrouter.toml on the next CI run instead of living in separate files that just wrap the OpenRouter endpoint. Removed: allenai, deepcogito, essentialai, inclusionai, inflection, liquid, meituan, nex-agi, nousresearch, prime-intellect, relace, switchpoint, tngtech, writer * fix(providers): remove 7 niche providers with no driver support No dedicated LLM driver code exists for these providers — they rely purely on OpenAI-compatible passthrough with no special handling. Removing them reduces registry noise; users can still reach them via openrouter.toml if needed. Removed: microsoft, ibm-granite, xiaomi, upstage, inception, aion-labs, arcee-ai * fix(providers): remove ai21, chutes, venice All three use ApiFormat::OpenAI with no special handling — pure passthrough. No registry entry needed; users can reach them via openrouter.toml or by adding a custom provider. * docs(providers): rewrite README with full provider catalog and inclusion criteria - List all 46 providers grouped by category with descriptions - Document why each provider exists (direct API, unique endpoint, dedicated driver, local, CLI) - Add inclusion criteria section explaining when to create standalone files vs merging into openrouter.toml - Document sync script routing logic - Update model counts: 49→46 providers, 339→232 models * docs: add comprehensive READMEs for all registry sections + deepinfra provider - agents/README.md: 32 agents across 7 categories with capability field reference - channels/README.md: 44 channels across 5 categories with protocol reference table - hands/README.md: 18 hands across 5 categories with HAND.toml format guide - mcp/README.md: 33 MCP servers across 5 categories with transport/auth format - plugins/README.md: 12 plugins with hook protocol documentation - skills/README.md: 60 skills across 9 categories with SKILL.md format guide - providers/deepinfra.toml: add DeepInfra serverless inference (5 models)
2026-04-24 00:02:33 +09:00
1 parent dbfb32d9d4
commit d43077afa9
@@ -1,16 +1,123 @@
 # Providers

-LLM provider and model metadata for LibreFang. Each provider file defines the provider's API configuration and all available models with pricing, context windows, and capability flags.
+LLM provider and model metadata for LibreFang. Each `.toml` file defines one provider's API configuration and all its available models with pricing, context windows, and capability flags.

-## Structure
+**Current state: 46 providers, 232+ models**

-```
-providers/
-├── anthropic.toml
-├── openai.toml
-├── groq.toml
-└── ...           (46 providers, 220+ models)
-```
+---
+
+## Provider Categories
+
+### Frontier / Major Cloud APIs
+
+| ID | Display Name | Base URL | API Key Env | Models | Description |
+|----|-------------|----------|-------------|--------|-------------|
+| `anthropic` | Anthropic | `https://api.anthropic.com` | `ANTHROPIC_API_KEY` | 7 | Claude family (Haiku / Sonnet / Opus); native Anthropic wire protocol, not OpenAI-compatible |
+| `openai` | OpenAI | `https://api.openai.com/v1` | `OPENAI_API_KEY` | 15 | GPT-4.x, o-series reasoning, image generation, TTS; also the fallback for codex-cli |
+| `gemini` | Google Gemini | `https://generativelanguage.googleapis.com` | `GEMINI_API_KEY` | 6 | Gemini 2.x family; native Google GenerativeLanguage protocol |
+| `xai` | xAI | `https://api.x.ai/v1` | `XAI_API_KEY` | 7 | Grok-3 family from Elon Musk's xAI |
+| `mistral` | Mistral AI | `https://api.mistral.ai/v1` | `MISTRAL_API_KEY` | 3 | Mistral Large / Small / Codestral; European frontier models |
+| `cohere` | Cohere | `https://api.cohere.com/v2` | `COHERE_API_KEY` | 4 | Command R+ family; strong RAG and tool-use models |
+| `deepseek` | DeepSeek | `https://api.deepseek.com/v1` | `DEEPSEEK_API_KEY` | 2 | DeepSeek-V3 (chat) + R1 (reasoning); extremely cost-effective |
+| `meta-llama` | Meta Llama | `https://api.llama.com/v1` | `LLAMA_API_KEY` | 4 | Official Meta Llama API — direct access to Llama 3.x |
+| `perplexity` | Perplexity AI | `https://api.perplexity.ai` | `PERPLEXITY_API_KEY` | 4 | Sonar family with live web search built in |
+
+### Fast Inference / Compute Clouds
+
+| ID | Display Name | Base URL | API Key Env | Models | Description |
+|----|-------------|----------|-------------|--------|-------------|
+| `groq` | Groq | `https://api.groq.com/openai/v1` | `GROQ_API_KEY` | 3 | GroqChip hardware; ~10× faster than GPU inference for supported models |
+| `cerebras` | Cerebras | `https://api.cerebras.ai/v1` | `CEREBRAS_API_KEY` | 4 | Wafer-scale chip inference; best-in-class throughput for Llama |
+| `sambanova` | SambaNova | `https://api.sambanova.ai/v1` | `SAMBANOVA_API_KEY` | 3 | Reconfigurable Dataflow Unit (RDU) inference; fast Llama variants |
+| `fireworks` | Fireworks AI | `https://api.fireworks.ai/inference/v1` | `FIREWORKS_API_KEY` | 5 | Serverless open-model hosting; fast cold-start |
+| `together` | Together AI | `https://api.together.xyz/v1` | `TOGETHER_API_KEY` | 8 | Open-model hosting (Llama, Qwen, Mistral) + fine-tuning API |
+| `nvidia-nim` | NVIDIA NIM | `https://integrate.api.nvidia.com/v1` | `NVIDIA_API_KEY` | 26 | NVIDIA NIM microservices; largest model selection in registry |
+| `replicate` | Replicate | `https://api.replicate.com/v1` | `REPLICATE_API_TOKEN` | 3 | Run any model as a serverless API; image + video + LLM |
+| `huggingface` | Hugging Face | `https://api-inference.huggingface.co/v1` | `HF_API_KEY` | 3 | HF Serverless Inference API for hosted open models |
+
+### Cloud Platform / Enterprise
+
+| ID | Display Name | Base URL | API Key Env | Models | Description |
+|----|-------------|----------|-------------|--------|-------------|
+| `bedrock` | AWS Bedrock | `https://bedrock-runtime.us-east-1.amazonaws.com` | `AWS_ACCESS_KEY_ID` | 11 | AWS-managed models (Claude, Llama, Mistral, Nova); IAM auth |
+| `vertex-ai` | Google Cloud Vertex AI | `https://us-central1-aiplatform.googleapis.com` | `GOOGLE_APPLICATION_CREDENTIALS` | 6 | GCP-hosted Gemini + third-party models; service account JSON auth |
+| `github-copilot` | GitHub Copilot | `https://api.githubcopilot.com` | `GITHUB_TOKEN` | 1 | Uses `ApiFormat::Copilot` — proprietary protocol, not OpenAI-compatible; requires GitHub PAT with Copilot access |
+
+### Aggregators / Routers
+
+| ID | Display Name | Base URL | API Key Env | Models | Description |
+|----|-------------|----------|-------------|--------|-------------|
+| `openrouter` | OpenRouter | `https://openrouter.ai/api/v1` | `OPENROUTER_API_KEY` | 18+ | Meta-provider routing to 300+ models; also receives models from OpenRouter-only providers via sync script |
+| `siliconflow` | SiliconFlow | `https://api.siliconflow.cn/v1` | `SILICONFLOW_API_KEY` | dynamic | 硅基流动 — Chinese open-model hosting; models discovered at runtime, not hardcoded in TOML |
+
+### Chinese Providers
+
+| ID | Display Name | Base URL | API Key Env | Models | Description |
+|----|-------------|----------|-------------|--------|-------------|
+| `qwen` | Qwen (Alibaba) | `https://dashscope.aliyuncs.com/compatible-mode/v1` | `DASHSCOPE_API_KEY` | 9 | Qwen3 family by Alibaba; multi-region support (`intl` / `us`) via `[provider.regions]` |
+| `moonshot` | Moonshot (Kimi) | `https://api.moonshot.ai/v1` | `MOONSHOT_API_KEY` | 5 | Kimi general chat models (moonshot-v1-8k/32k/128k) |
+| `minimax` | MiniMax | `https://api.minimax.io/v1` | `MINIMAX_API_KEY` | 8 | MiniMax M-series; strong Chinese + multilingual models |
+| `zhipu` | Zhipu AI (GLM) | `https://open.bigmodel.cn/api/paas/v4` | `ZHIPU_API_KEY` | 6 | GLM-4 family by Zhipu AI (智谱); general chat + vision |
+| `zai` | Z.AI | `https://api.z.ai/api/paas/v4` | `ZHIPU_API_KEY` | 2 | Z.AI general models; shares API key with zhipu |
+| `baichuan` | Baichuan (百川) | `https://api.baichuan-ai.com/v1` | `BAICHUAN_API_KEY` | 2 | Baichuan4 family; strong Chinese-language performance |
+| `volcengine` | Volcano Engine (Doubao) | `https://ark.cn-beijing.volces.com/api/v3` | `VOLCENGINE_API_KEY` | 8 | ByteDance Doubao models; Ark platform |
+| `stepfun` | Stepfun (阶跃星辰) | `https://api.stepfun.com/v1` | `STEPFUN_API_KEY` | 4 | Step-2 family; strong long-context and reasoning |
+| `tencent` | Tencent | `https://api.hunyuan.cloud.tencent.com/v1` | `HUNYUAN_API_KEY` | 1 | Hunyuan models by Tencent |
+| `qianfan` | Baidu Qianfan | `https://qianfan.baidubce.com/v2` | `QIANFAN_API_KEY` | 3 | ERNIE family by Baidu; Qianfan platform |
+
+### Coding-Specific Endpoints
+
+Separate `base_url` for coding workloads — not just model aliases. Same API key as the general counterpart.
+
+| ID | Display Name | Base URL | API Key Env | Models | vs. General |
+|----|-------------|----------|-------------|--------|-------------|
+| `kimi_coding` | Kimi for Code | `https://api.kimi.com/coding` | `KIMI_API_KEY` | 1 | vs `moonshot`: `api.moonshot.ai/v1` |
+| `alibaba-coding-plan` | Alibaba Coding Plan (Intl) | `https://coding-intl.dashscope.aliyuncs.com/v1` | `ALIBABA_CODING_PLAN_API_KEY` | 9 | vs `qwen`: `dashscope.aliyuncs.com`; aggregates models from multiple vendors (MiniMax, GLM, Kimi) |
+| `volcengine_coding` | Volcano Engine Coding Plan | `https://ark.cn-beijing.volces.com/api/coding/v3` | `VOLCENGINE_API_KEY` | dynamic | vs `volcengine`: `/api/v3`; models discovered at runtime |
+| `zhipu_coding` | Zhipu Coding (CodeGeeX) | `https://open.bigmodel.cn/api/coding/paas/v4` | `ZHIPU_API_KEY` | 1 | vs `zhipu`: `/api/paas/v4`; CodeGeeX-4 coding model |
+| `zai_coding` | Z.AI Coding | `https://api.z.ai/api/coding/paas/v4` | `ZHIPU_API_KEY` | 2 | vs `zai`: `/api/paas/v4`; GLM coding variants |
+
+### CLI-Based Providers (No API Key)
+
+Route through a locally-installed CLI tool. `key_required = false`, `base_url` is empty. Cost is $0.
+
+| ID | Display Name | CLI Binary | Models | ApiFormat | Description |
+|----|-------------|-----------|--------|-----------|-------------|
+| `claude-code` | Claude Code | `claude` | 3 | `ClaudeCode` | Anthropic's official CLI; routes through Claude API with OAuth |
+| `codex-cli` | Codex CLI | `codex` | 6 | `CodexCli` | OpenAI Codex CLI; also serves as fallback for `openai` provider |
+| `gemini-cli` | Gemini CLI | `gemini` | 2 | `GeminiCli` | Google Gemini CLI; also serves as fallback for `gemini` provider |
+| `qwen-code` | Qwen Code | `qwen-code` | 3 | `QwenCode` | Alibaba Qwen coding CLI |
+
+### Local / Self-Hosted
+
+| ID | Display Name | Default Base URL | API Key Env | Notes |
+|----|-------------|-----------------|-------------|-------|
+| `ollama` | Ollama | `http://localhost:11434/v1` | `OLLAMA_API_KEY` | No key required; models discovered dynamically at runtime via `/api/tags` |
+| `lmstudio` | LM Studio | `http://localhost:1234/v1` | `LMSTUDIO_API_KEY` | No key required; GUI app for running GGUF models locally |
+| `vllm` | vLLM | `http://localhost:8000/v1` | `VLLM_API_KEY` | No key required; high-throughput inference server for production self-hosting |
+
+### Special / Niche
+
+| ID | Display Name | Base URL | API Key Env | Notes |
+|----|-------------|----------|-------------|-------|
+| `chatgpt` | ChatGPT (Session Auth) | `https://chatgpt.com/backend-api` | `CHATGPT_SESSION_TOKEN` | Session cookie auth, not an API key. Exposes GPT-5.x Codex models (gpt-5.1-codex etc.) that are unavailable via the standard OpenAI API |
+| `elevenlabs` | ElevenLabs | `https://api.elevenlabs.io/v1` | `ELEVENLABS_API_KEY` | TTS / voice generation only — has a dedicated `elevenlabs.rs` driver in librefang-runtime. No chat models; appears in the provider list for media capability routing |
+
+---
+
+## Inclusion Criteria
+
+A provider gets its own `.toml` file when it meets **at least one** of:
+
+1. Has a **direct public API** not accessible via OpenRouter
+2. Has a **unique endpoint** for a specific workload (e.g. coding plan endpoints)
+3. Has a **dedicated driver** (`ApiFormat` beyond generic `OpenAI`)
+4. Is a **local/self-hosted** runtime
+5. Is a **CLI-based** provider
+
+Providers that only route through `openrouter.ai/api/v1` with `OPENROUTER_API_KEY` are **not** given standalone files — their models are merged into `openrouter.toml` by the sync script. See [scripts/sync-pricing.py](../scripts/sync-pricing.py).
+
+---

 ## Provider TOML Format

@@ -39,26 +146,52 @@ aliases = ["short-name"]
 ## Tier Definitions

 | Tier | Description | Examples |
-|------|-------------|----------|
+|------|-------------|---------|
 | `frontier` | Most capable, cutting-edge | Claude Opus, GPT-4.1 |
 | `smart` | Smart, cost-effective | Claude Sonnet, Gemini 2.5 Flash |
 | `balanced` | Balanced speed/cost | GPT-4.1 Mini, Llama 3.3 70B |
 | `fast` | Fastest, cheapest | GPT-4o Mini, Claude Haiku |
 | `local` | Local models, zero cost | Ollama, vLLM, LM Studio |

+---
+
+## Sync Script
+
+`scripts/sync-pricing.py` runs daily via CI to keep model pricing current.
+
+```bash
+python scripts/sync-pricing.py                             # Update prices only
+python scripts/sync-pricing.py --create-missing            # Also add new providers
+python scripts/sync-pricing.py --dry-run --create-missing  # Preview changes
+```
+
+**`--create-missing` routing logic:**
+
+| Condition | Action |
+|-----------|--------|
+| Provider in `PROVIDER_API` map (has direct API) | Create standalone `.toml` |
+| Provider not in `PROVIDER_API` (OpenRouter-only) | Merge into `openrouter.toml` with `openrouter/{provider}/{model}` IDs |
+| Provider in `SKIP_PROVIDERS` (morph, aider, kwaipilot, …) | Skip entirely |
+| Provider ID starts with `~` (OpenRouter internal routing alias) | Skip entirely |
+
+---
+
 ## Validation

 ```bash
-python scripts/validate.py
+python scripts/validate.py           # Warn on issues
+python scripts/validate.py --strict  # Treat warnings as errors
 ```

 Checks: required fields, valid tiers, non-negative costs, no duplicate model IDs.

-## Adding or Updating a Model
+## Adding or Updating a Provider

-1. Edit or create the provider file in `providers/`
-2. Use exact API model IDs and verify pricing from official sources
-3. Run `python scripts/validate.py`
-4. Submit a PR
+1. Check inclusion criteria above — if OpenRouter-only, don't create a new file
+2. Create or edit the `.toml` in `providers/`
+3. Use exact API model IDs; verify pricing from official sources
+4. Run `python scripts/validate.py`
+5. Update the table in this README
+6. Submit a PR

-See [CONTRIBUTING.md](../CONTRIBUTING.md) for the full guide and pricing source links.
+See [CONTRIBUTING.md](../CONTRIBUTING.md) for the full guide.
@@ -1,48 +0,0 @@
-# AI21 Labs — https://ai21.com
-# Models: 3
-
-[provider]
-id = "ai21"
-display_name = "AI21 Labs"
-api_key_env = "AI21_API_KEY"
-base_url = "https://api.ai21.com/studio/v1"
-key_required = true
-
-[[models]]
-id = "jamba-1.5-large"
-display_name = "Jamba 1.5 Large"
-tier = "smart"
-context_window = 256000
-max_output_tokens = 4096
-input_cost_per_m = 2.0
-output_cost_per_m = 8.0
-supports_tools = true
-supports_vision = false
-supports_streaming = true
-aliases = ["jamba"]
-
-[[models]]
-id = "jamba-1.5-mini"
-display_name = "Jamba 1.5 Mini"
-tier = "fast"
-context_window = 256000
-max_output_tokens = 4096
-input_cost_per_m = 0.20
-output_cost_per_m = 0.40
-supports_tools = true
-supports_vision = false
-supports_streaming = true
-aliases = []
-
-[[models]]
-id = "jamba-instruct"
-display_name = "Jamba Instruct"
-tier = "balanced"
-context_window = 256000
-max_output_tokens = 4096
-input_cost_per_m = 0.50
-output_cost_per_m = 0.70
-supports_tools = true
-supports_vision = false
-supports_streaming = true
-aliases = []
@@ -1,35 +0,0 @@
-[provider]
-id = "aider"
-display_name = "Aider"
-api_key_env = ""
-base_url = ""
-key_required = false
-
-[[models]]
-id = "aider/sonnet"
-display_name = "Claude Sonnet via Aider"
-provider = "aider"
-tier = "smart"
-context_window = 200000
-max_output_tokens = 64000
-input_cost_per_m = 0.0
-output_cost_per_m = 0.0
-supports_tools = false
-supports_vision = false
-supports_streaming = false
-supports_thinking = true
-aliases = ["aider-sonnet"]
-
-[[models]]
-id = "aider/gpt-4o"
-display_name = "GPT-4o via Aider"
-provider = "aider"
-tier = "smart"
-context_window = 128000
-max_output_tokens = 16384
-input_cost_per_m = 0.0
-output_cost_per_m = 0.0
-supports_tools = false
-supports_vision = false
-supports_streaming = false
-aliases = ["aider-gpt-4o"]
@@ -1,48 +0,0 @@
-# aion-labs — auto-generated from OpenRouter API
-
-[provider]
-id = "aion-labs"
-display_name = "Aion Labs"
-api_key_env = "AION_LABS_API_KEY"
-base_url = "https://api.aionlabs.ai/v1"
-key_required = true
-
-[[models]]
-id = "aion-1.0"
-display_name = "AionLabs: Aion-1.0"
-tier = "frontier"
-context_window = 131072
-max_output_tokens = 32768
-input_cost_per_m = 4.0
-output_cost_per_m = 8.0
-supports_streaming = true
-
-[[models]]
-id = "aion-1.0-mini"
-display_name = "AionLabs: Aion-1.0-Mini"
-tier = "smart"
-context_window = 131072
-max_output_tokens = 32768
-input_cost_per_m = 0.7
-output_cost_per_m = 1.4
-supports_streaming = true
-
-[[models]]
-id = "aion-2.0"
-display_name = "AionLabs: Aion-2.0"
-tier = "smart"
-context_window = 131072
-max_output_tokens = 32768
-input_cost_per_m = 0.8
-output_cost_per_m = 1.6
-supports_streaming = true
-
-[[models]]
-id = "aion-rp-llama-3.1-8b"
-display_name = "AionLabs: Aion-RP 1.0 (8B)"
-tier = "smart"
-context_window = 32768
-max_output_tokens = 32768
-input_cost_per_m = 0.8
-output_cost_per_m = 1.6
-supports_streaming = true
@@ -1,50 +0,0 @@
-# allenai — auto-generated from OpenRouter API
-
-[provider]
-id = "allenai"
-display_name = "Allenai"
-api_key_env = "OPENROUTER_API_KEY"
-base_url = "https://openrouter.ai/api/v1"
-key_required = true
-
-[[models]]
-id = "olmo-2-0325-32b-instruct"
-display_name = "AllenAI: Olmo 2 32B Instruct"
-tier = "fast"
-context_window = 128000
-max_output_tokens = 16384
-input_cost_per_m = 0.05
-output_cost_per_m = 0.2
-supports_streaming = true
-
-[[models]]
-id = "olmo-3-32b-think"
-display_name = "AllenAI: Olmo 3 32B Think"
-tier = "fast"
-context_window = 65536
-max_output_tokens = 65536
-input_cost_per_m = 0.15
-output_cost_per_m = 0.5
-supports_streaming = true
-supports_thinking = true
-
-[[models]]
-id = "olmo-3.1-32b-instruct"
-display_name = "AllenAI: Olmo 3.1 32B Instruct"
-tier = "fast"
-context_window = 65536
-max_output_tokens = 16384
-input_cost_per_m = 0.2
-output_cost_per_m = 0.6
-supports_streaming = true
-
-[[models]]
-id = "olmo-3.1-32b-think"
-display_name = "AllenAI: Olmo 3.1 32B Think"
-tier = "fast"
-context_window = 65536
-max_output_tokens = 65536
-input_cost_per_m = 0.15
-output_cost_per_m = 0.5
-supports_streaming = true
-supports_thinking = true
@@ -1,78 +0,0 @@
-# arcee-ai — auto-generated from OpenRouter API
-
-[provider]
-id = "arcee-ai"
-display_name = "Arcee Ai"
-api_key_env = "ARCEE_API_KEY"
-base_url = "https://api.arcee.ai/v1"
-key_required = true
-
-[[models]]
-id = "coder-large"
-display_name = "Arcee AI: Coder Large"
-tier = "smart"
-context_window = 32768
-max_output_tokens = 8192
-input_cost_per_m = 0.5
-output_cost_per_m = 0.8
-supports_streaming = true
-
-[[models]]
-id = "maestro-reasoning"
-display_name = "Arcee AI: Maestro Reasoning"
-tier = "smart"
-context_window = 131072
-max_output_tokens = 32000
-input_cost_per_m = 0.9
-output_cost_per_m = 3.3
-supports_streaming = true
-
-[[models]]
-id = "spotlight"
-display_name = "Arcee AI: Spotlight"
-tier = "fast"
-context_window = 131072
-max_output_tokens = 65537
-input_cost_per_m = 0.18
-output_cost_per_m = 0.18
-supports_streaming = true
-
-[[models]]
-id = "trinity-large-preview:free"
-display_name = "Arcee AI: Trinity Large Preview (free)"
-tier = "fast"
-context_window = 131000
-max_output_tokens = 16384
-input_cost_per_m = 0.0
-output_cost_per_m = 0.0
-supports_streaming = true
-
-[[models]]
-id = "trinity-large-thinking"
-display_name = "Arcee AI: Trinity Large Thinking"
-tier = "fast"
-context_window = 262144
-max_output_tokens = 262144
-input_cost_per_m = 0.22
-output_cost_per_m = 0.85
-supports_streaming = true
-
-[[models]]
-id = "trinity-mini"
-display_name = "Arcee AI: Trinity Mini"
-tier = "fast"
-context_window = 131072
-max_output_tokens = 131072
-input_cost_per_m = 0.045
-output_cost_per_m = 0.15
-supports_streaming = true
-
-[[models]]
-id = "virtuoso-large"
-display_name = "Arcee AI: Virtuoso Large"
-tier = "smart"
-context_window = 131072
-max_output_tokens = 64000
-input_cost_per_m = 0.75
-output_cost_per_m = 1.2
-supports_streaming = true
@@ -1,76 +0,0 @@
-# Chutes.ai — https://chutes.ai
-# Models: 5
-
-[provider]
-id = "chutes"
-display_name = "Chutes.ai"
-api_key_env = "CHUTES_API_KEY"
-base_url = "https://llm.chutes.ai/v1"
-key_required = true
-
-[[models]]
-id = "chutes/deepseek-ai/DeepSeek-V3"
-display_name = "DeepSeek V3 (Chutes)"
-tier = "smart"
-context_window = 128000
-max_output_tokens = 8192
-input_cost_per_m = 0.25
-output_cost_per_m = 0.35
-supports_tools = true
-supports_vision = false
-supports_streaming = true
-aliases = ["chutes-deepseek-v3"]
-
-[[models]]
-id = "chutes/deepseek-ai/DeepSeek-R1"
-display_name = "DeepSeek R1 (Chutes)"
-tier = "smart"
-context_window = 128000
-max_output_tokens = 8192
-input_cost_per_m = 0.55
-output_cost_per_m = 2.19
-supports_tools = false
-supports_vision = false
-supports_streaming = true
-supports_thinking = true
-aliases = ["chutes-deepseek-r1"]
-
-[[models]]
-id = "chutes/meta-llama/Llama-4-Maverick-17B-128E-Instruct"
-display_name = "Llama 4 Maverick (Chutes)"
-tier = "balanced"
-context_window = 128000
-max_output_tokens = 8192
-input_cost_per_m = 0.20
-output_cost_per_m = 0.30
-supports_tools = true
-supports_vision = false
-supports_streaming = true
-aliases = ["chutes-llama-maverick"]
-
-[[models]]
-id = "chutes/Qwen/Qwen3-235B-A22B"
-display_name = "Qwen3 235B (Chutes)"
-tier = "smart"
-context_window = 128000
-max_output_tokens = 8192
-input_cost_per_m = 0.25
-output_cost_per_m = 0.35
-supports_tools = true
-supports_vision = false
-supports_streaming = true
-supports_thinking = true
-aliases = ["chutes-qwen3"]
-
-[[models]]
-id = "chutes/meta-llama/Llama-3.3-70B-Instruct"
-display_name = "Llama 3.3 70B (Chutes)"
-tier = "balanced"
-context_window = 128000
-max_output_tokens = 8192
-input_cost_per_m = 0.10
-output_cost_per_m = 0.15
-supports_tools = true
-supports_vision = false
-supports_streaming = true
-aliases = ["chutes-llama-70b"]
@@ -1,18 +0,0 @@
-# deepcogito — auto-generated from OpenRouter API
-
-[provider]
-id = "deepcogito"
-display_name = "Deepcogito"
-api_key_env = "OPENROUTER_API_KEY"
-base_url = "https://openrouter.ai/api/v1"
-key_required = true
-
-[[models]]
-id = "cogito-v2.1-671b"
-display_name = "Deep Cogito: Cogito v2.1 671B"
-tier = "smart"
-context_window = 128000
-max_output_tokens = 16384
-input_cost_per_m = 1.25
-output_cost_per_m = 1.25
-supports_streaming = true
@@ -0,0 +1,69 @@
+# DeepInfra — https://deepinfra.com
+# Serverless open-model inference (Llama, Mistral, Qwen, etc.)
+
+[provider]
+id = "deepinfra"
+display_name = "DeepInfra"
+api_key_env = "DEEPINFRA_API_KEY"
+base_url = "https://api.deepinfra.com/v1/openai"
+key_required = true
+
+[[models]]
+id = "meta-llama/Meta-Llama-3.1-8B-Instruct"
+display_name = "Llama 3.1 8B Instruct"
+tier = "fast"
+context_window = 128000
+max_output_tokens = 8192
+input_cost_per_m = 0.06
+output_cost_per_m = 0.06
+supports_tools = true
+supports_vision = false
+supports_streaming = true
+
+[[models]]
+id = "meta-llama/Meta-Llama-3.1-70B-Instruct"
+display_name = "Llama 3.1 70B Instruct"
+tier = "balanced"
+context_window = 128000
+max_output_tokens = 8192
+input_cost_per_m = 0.35
+output_cost_per_m = 0.4
+supports_tools = true
+supports_vision = false
+supports_streaming = true
+
+[[models]]
+id = "meta-llama/Meta-Llama-3.1-405B-Instruct"
+display_name = "Llama 3.1 405B Instruct"
+tier = "smart"
+context_window = 128000
+max_output_tokens = 8192
+input_cost_per_m = 0.8
+output_cost_per_m = 0.8
+supports_tools = true
+supports_vision = false
+supports_streaming = true
+
+[[models]]
+id = "mistralai/Mistral-7B-Instruct-v0.3"
+display_name = "Mistral 7B Instruct v0.3"
+tier = "fast"
+context_window = 32768
+max_output_tokens = 8192
+input_cost_per_m = 0.07
+output_cost_per_m = 0.07
+supports_tools = true
+supports_vision = false
+supports_streaming = true
+
+[[models]]
+id = "Qwen/Qwen2.5-72B-Instruct"
+display_name = "Qwen 2.5 72B Instruct"
+tier = "balanced"
+context_window = 32768
+max_output_tokens = 8192
+input_cost_per_m = 0.35
+output_cost_per_m = 0.4
+supports_tools = true
+supports_vision = false
+supports_streaming = true
@@ -1,18 +0,0 @@
-# essentialai — auto-generated from OpenRouter API
-
-[provider]
-id = "essentialai"
-display_name = "Essentialai"
-api_key_env = "OPENROUTER_API_KEY"
-base_url = "https://openrouter.ai/api/v1"
-key_required = true
-
-[[models]]
-id = "rnj-1-instruct"
-display_name = "EssentialAI: Rnj 1 Instruct"
-tier = "fast"
-context_window = 32768
-max_output_tokens = 8192
-input_cost_per_m = 0.15
-output_cost_per_m = 0.15
-supports_streaming = true
@@ -1,18 +0,0 @@
-# ibm-granite — auto-generated from OpenRouter API
-
-[provider]
-id = "ibm-granite"
-display_name = "Ibm Granite"
-api_key_env = "WATSONX_API_KEY"
-base_url = "https://us-south.ml.cloud.ibm.com/ml/v1"
-key_required = true
-
-[[models]]
-id = "granite-4.0-h-micro"
-display_name = "IBM: Granite 4.0 Micro"
-tier = "fast"
-context_window = 131000
-max_output_tokens = 16384
-input_cost_per_m = 0.017
-output_cost_per_m = 0.11
-supports_streaming = true
@@ -1,18 +0,0 @@
-# inception — auto-generated from OpenRouter API
-
-[provider]
-id = "inception"
-display_name = "Inception"
-api_key_env = "INCEPTION_API_KEY"
-base_url = "https://api.inceptionlabs.ai/v1"
-key_required = true
-
-[[models]]
-id = "mercury-2"
-display_name = "Inception: Mercury 2"
-tier = "fast"
-context_window = 128000
-max_output_tokens = 50000
-input_cost_per_m = 0.25
-output_cost_per_m = 0.75
-supports_streaming = true
@@ -1,18 +0,0 @@
-# inclusionai — auto-generated from OpenRouter API
-
-[provider]
-id = "inclusionai"
-display_name = "Inclusionai"
-api_key_env = "OPENROUTER_API_KEY"
-base_url = "https://openrouter.ai/api/v1"
-key_required = true
-
-[[models]]
-id = "ling-2.6-flash:free"
-display_name = "inclusionAI: Ling-2.6-flash (free)"
-tier = "fast"
-context_window = 262144
-max_output_tokens = 32768
-input_cost_per_m = 0.0
-output_cost_per_m = 0.0
-supports_streaming = true
@@ -1,28 +0,0 @@
-# inflection — auto-generated from OpenRouter API
-
-[provider]
-id = "inflection"
-display_name = "Inflection"
-api_key_env = "OPENROUTER_API_KEY"
-base_url = "https://openrouter.ai/api/v1"
-key_required = true
-
-[[models]]
-id = "inflection-3-pi"
-display_name = "Inflection: Inflection 3 Pi"
-tier = "smart"
-context_window = 8000
-max_output_tokens = 1024
-input_cost_per_m = 2.5
-output_cost_per_m = 10.0
-supports_streaming = true
-
-[[models]]
-id = "inflection-3-productivity"
-display_name = "Inflection: Inflection 3 Productivity"
-tier = "smart"
-context_window = 8000
-max_output_tokens = 1024
-input_cost_per_m = 2.5
-output_cost_per_m = 10.0
-supports_streaming = true
@@ -1,18 +0,0 @@
-# kwaipilot — auto-generated from OpenRouter API
-
-[provider]
-id = "kwaipilot"
-display_name = "Kwaipilot"
-api_key_env = "OPENROUTER_API_KEY"
-base_url = "https://openrouter.ai/api/v1"
-key_required = true
-
-[[models]]
-id = "kat-coder-pro-v2"
-display_name = "Kwaipilot: KAT-Coder-Pro V2"
-tier = "fast"
-context_window = 256000
-max_output_tokens = 80000
-input_cost_per_m = 0.3
-output_cost_per_m = 1.2
-supports_streaming = true
@@ -1,38 +0,0 @@
-# liquid — auto-generated from OpenRouter API
-
-[provider]
-id = "liquid"
-display_name = "Liquid"
-api_key_env = "OPENROUTER_API_KEY"
-base_url = "https://openrouter.ai/api/v1"
-key_required = true
-
-[[models]]
-id = "lfm-2-24b-a2b"
-display_name = "LiquidAI: LFM2-24B-A2B"
-tier = "fast"
-context_window = 32768
-max_output_tokens = 8192
-input_cost_per_m = 0.03
-output_cost_per_m = 0.12
-supports_streaming = true
-
-[[models]]
-id = "lfm-2.5-1.2b-instruct:free"
-display_name = "LiquidAI: LFM2.5-1.2B-Instruct (free)"
-tier = "fast"
-context_window = 32768
-max_output_tokens = 8192
-input_cost_per_m = 0.0
-output_cost_per_m = 0.0
-supports_streaming = true
-
-[[models]]
-id = "lfm-2.5-1.2b-thinking:free"
-display_name = "LiquidAI: LFM2.5-1.2B-Thinking (free)"
-tier = "fast"
-context_window = 32768
-max_output_tokens = 8192
-input_cost_per_m = 0.0
-output_cost_per_m = 0.0
-supports_streaming = true
@@ -1,18 +0,0 @@
-# meituan — auto-generated from OpenRouter API
-
-[provider]
-id = "meituan"
-display_name = "Meituan"
-api_key_env = "OPENROUTER_API_KEY"
-base_url = "https://openrouter.ai/api/v1"
-key_required = true
-
-[[models]]
-id = "longcat-flash-chat"
-display_name = "Meituan: LongCat Flash Chat"
-tier = "fast"
-context_window = 131072
-max_output_tokens = 131072
-input_cost_per_m = 0.2
-output_cost_per_m = 0.8
-supports_streaming = true
@@ -1,28 +0,0 @@
-# microsoft — auto-generated from OpenRouter API
-
-[provider]
-id = "microsoft"
-display_name = "Microsoft"
-api_key_env = "GITHUB_TOKEN"
-base_url = "https://models.inference.ai.azure.com"
-key_required = true
-
-[[models]]
-id = "phi-4"
-display_name = "Microsoft: Phi 4"
-tier = "fast"
-context_window = 16384
-max_output_tokens = 16384
-input_cost_per_m = 0.065
-output_cost_per_m = 0.14
-supports_streaming = true
-
-[[models]]
-id = "wizardlm-2-8x22b"
-display_name = "WizardLM-2 8x22B"
-tier = "smart"
-context_window = 65535
-max_output_tokens = 8000
-input_cost_per_m = 0.62
-output_cost_per_m = 0.62
-supports_streaming = true
@@ -1,28 +0,0 @@
-# morph — auto-generated from OpenRouter API
-
-[provider]
-id = "morph"
-display_name = "Morph"
-api_key_env = "MORPH_API_KEY"
-base_url = "https://api.morphllm.com/v1"
-key_required = true
-
-[[models]]
-id = "morph-v3-fast"
-display_name = "Morph: Morph V3 Fast"
-tier = "smart"
-context_window = 81920
-max_output_tokens = 38000
-input_cost_per_m = 0.8
-output_cost_per_m = 1.2
-supports_streaming = true
-
-[[models]]
-id = "morph-v3-large"
-display_name = "Morph: Morph V3 Large"
-tier = "smart"
-context_window = 262144
-max_output_tokens = 131072
-input_cost_per_m = 0.9
-output_cost_per_m = 1.9
-supports_streaming = true
@@ -1,18 +0,0 @@
-# nex-agi — auto-generated from OpenRouter API
-
-[provider]
-id = "nex-agi"
-display_name = "Nex Agi"
-api_key_env = "OPENROUTER_API_KEY"
-base_url = "https://openrouter.ai/api/v1"
-key_required = true
-
-[[models]]
-id = "deepseek-v3.1-nex-n1"
-display_name = "Nex AGI: DeepSeek V3.1 Nex N1"
-tier = "fast"
-context_window = 131072
-max_output_tokens = 163840
-input_cost_per_m = 0.135
-output_cost_per_m = 0.5
-supports_streaming = true
@@ -1,68 +0,0 @@
-# nousresearch — auto-generated from OpenRouter API
-
-[provider]
-id = "nousresearch"
-display_name = "Nousresearch"
-api_key_env = "OPENROUTER_API_KEY"
-base_url = "https://openrouter.ai/api/v1"
-key_required = true
-
-[[models]]
-id = "hermes-2-pro-llama-3-8b"
-display_name = "NousResearch: Hermes 2 Pro - Llama-3 8B"
-tier = "fast"
-context_window = 8192
-max_output_tokens = 8192
-input_cost_per_m = 0.14
-output_cost_per_m = 0.14
-supports_streaming = true
-
-[[models]]
-id = "hermes-3-llama-3.1-405b"
-display_name = "Nous: Hermes 3 405B Instruct"
-tier = "smart"
-context_window = 131072
-max_output_tokens = 16384
-input_cost_per_m = 1.0
-output_cost_per_m = 1.0
-supports_streaming = true
-
-[[models]]
-id = "hermes-3-llama-3.1-405b:free"
-display_name = "Nous: Hermes 3 405B Instruct (free)"
-tier = "fast"
-context_window = 131072
-max_output_tokens = 16384
-input_cost_per_m = 0.0
-output_cost_per_m = 0.0
-supports_streaming = true
-
-[[models]]
-id = "hermes-3-llama-3.1-70b"
-display_name = "Nous: Hermes 3 70B Instruct"
-tier = "fast"
-context_window = 131072
-max_output_tokens = 16384
-input_cost_per_m = 0.3
-output_cost_per_m = 0.3
-supports_streaming = true
-
-[[models]]
-id = "hermes-4-405b"
-display_name = "Nous: Hermes 4 405B"
-tier = "smart"
-context_window = 131072
-max_output_tokens = 16384
-input_cost_per_m = 1.0
-output_cost_per_m = 3.0
-supports_streaming = true
-
-[[models]]
-id = "hermes-4-70b"
-display_name = "Nous: Hermes 4 70B"
-tier = "fast"
-context_window = 131072
-max_output_tokens = 16384
-input_cost_per_m = 0.13
-output_cost_per_m = 0.4
-supports_streaming = true
@@ -1,18 +0,0 @@
-# prime-intellect — auto-generated from OpenRouter API
-
-[provider]
-id = "prime-intellect"
-display_name = "Prime Intellect"
-api_key_env = "OPENROUTER_API_KEY"
-base_url = "https://openrouter.ai/api/v1"
-key_required = true
-
-[[models]]
-id = "intellect-3"
-display_name = "Prime Intellect: INTELLECT-3"
-tier = "fast"
-context_window = 131072
-max_output_tokens = 131072
-input_cost_per_m = 0.2
-output_cost_per_m = 1.1
-supports_streaming = true
@@ -1,28 +0,0 @@
-# relace — auto-generated from OpenRouter API
-
-[provider]
-id = "relace"
-display_name = "Relace"
-api_key_env = "OPENROUTER_API_KEY"
-base_url = "https://openrouter.ai/api/v1"
-key_required = true
-
-[[models]]
-id = "relace-apply-3"
-display_name = "Relace: Relace Apply 3"
-tier = "smart"
-context_window = 256000
-max_output_tokens = 128000
-input_cost_per_m = 0.85
-output_cost_per_m = 1.25
-supports_streaming = true
-
-[[models]]
-id = "relace-search"
-display_name = "Relace: Relace Search"
-tier = "smart"
-context_window = 256000
-max_output_tokens = 128000
-input_cost_per_m = 1.0
-output_cost_per_m = 3.0
-supports_streaming = true
@@ -1,18 +0,0 @@
-# switchpoint — auto-generated from OpenRouter API
-
-[provider]
-id = "switchpoint"
-display_name = "Switchpoint"
-api_key_env = "OPENROUTER_API_KEY"
-base_url = "https://openrouter.ai/api/v1"
-key_required = true
-
-[[models]]
-id = "router"
-display_name = "Switchpoint Router"
-tier = "smart"
-context_window = 131072
-max_output_tokens = 16384
-input_cost_per_m = 0.85
-output_cost_per_m = 3.4
-supports_streaming = true
@@ -1,18 +0,0 @@
-# tngtech — auto-generated from OpenRouter API
-
-[provider]
-id = "tngtech"
-display_name = "Tngtech"
-api_key_env = "OPENROUTER_API_KEY"
-base_url = "https://openrouter.ai/api/v1"
-key_required = true
-
-[[models]]
-id = "deepseek-r1t2-chimera"
-display_name = "TNG: DeepSeek R1T2 Chimera"
-tier = "fast"
-context_window = 163840
-max_output_tokens = 163840
-input_cost_per_m = 0.3
-output_cost_per_m = 1.1
-supports_streaming = true
@@ -1,18 +0,0 @@
-# upstage — auto-generated from OpenRouter API
-
-[provider]
-id = "upstage"
-display_name = "Upstage"
-api_key_env = "UPSTAGE_API_KEY"
-base_url = "https://api.upstage.ai/v1"
-key_required = true
-
-[[models]]
-id = "solar-pro-3"
-display_name = "Upstage: Solar Pro 3"
-tier = "fast"
-context_window = 128000
-max_output_tokens = 16384
-input_cost_per_m = 0.15
-output_cost_per_m = 0.6
-supports_streaming = true
@@ -1,49 +0,0 @@
-# Venice.ai — https://venice.ai
-# Models: 3
-
-[provider]
-id = "venice"
-display_name = "Venice.ai"
-api_key_env = "VENICE_API_KEY"
-base_url = "https://api.venice.ai/api/v1"
-key_required = true
-
-[[models]]
-id = "venice-uncensored"
-display_name = "Venice Uncensored"
-tier = "fast"
-context_window = 32000
-max_output_tokens = 8192
-input_cost_per_m = 0.20
-output_cost_per_m = 0.90
-supports_tools = true
-supports_vision = false
-supports_streaming = true
-aliases = ["venice"]
-
-[[models]]
-id = "llama-3.3-70b"
-display_name = "Llama 3.3 70B (Venice)"
-tier = "balanced"
-context_window = 128000
-max_output_tokens = 8192
-input_cost_per_m = 0.1
-output_cost_per_m = 0.32
-supports_tools = true
-supports_vision = false
-supports_streaming = true
-aliases = []
-
-[[models]]
-id = "qwen3-235b-a22b-instruct-2507"
-display_name = "Qwen3 235B A22B (Venice)"
-tier = "smart"
-context_window = 128000
-max_output_tokens = 8192
-input_cost_per_m = 0.20
-output_cost_per_m = 0.90
-supports_tools = true
-supports_vision = false
-supports_streaming = true
-supports_thinking = true
-aliases = []
@@ -1,18 +0,0 @@
-# writer — auto-generated from OpenRouter API
-
-[provider]
-id = "writer"
-display_name = "Writer"
-api_key_env = "OPENROUTER_API_KEY"
-base_url = "https://openrouter.ai/api/v1"
-key_required = true
-
-[[models]]
-id = "palmyra-x5"
-display_name = "Writer: Palmyra X5"
-tier = "smart"
-context_window = 1040000
-max_output_tokens = 8192
-input_cost_per_m = 0.6
-output_cost_per_m = 6.0
-supports_streaming = true
@@ -1,38 +0,0 @@
-# xiaomi — auto-generated from OpenRouter API
-
-[provider]
-id = "xiaomi"
-display_name = "Xiaomi"
-api_key_env = "XIAOMI_ACCESS_KEY_ID"
-base_url = "https://cnbj3-cloud-ml.api.xiaomi.net"
-key_required = true
-
-[[models]]
-id = "mimo-v2-flash"
-display_name = "Xiaomi: MiMo-V2-Flash"
-tier = "fast"
-context_window = 262144
-max_output_tokens = 65536
-input_cost_per_m = 0.09
-output_cost_per_m = 0.29
-supports_streaming = true
-
-[[models]]
-id = "mimo-v2-omni"
-display_name = "Xiaomi: MiMo-V2-Omni"
-tier = "fast"
-context_window = 262144
-max_output_tokens = 65536
-input_cost_per_m = 0.4
-output_cost_per_m = 2.0
-supports_streaming = true
-
-[[models]]
-id = "mimo-v2-pro"
-display_name = "Xiaomi: MiMo-V2-Pro"
-tier = "smart"
-context_window = 1048576
-max_output_tokens = 131072
-input_cost_per_m = 1.0
-output_cost_per_m = 3.0
-supports_streaming = true
@@ -1,18 +0,0 @@
-# ~anthropic — auto-generated from OpenRouter API
-
-[provider]
-id = "~anthropic"
-display_name = "~Anthropic"
-api_key_env = "OPENROUTER_API_KEY"
-base_url = "https://openrouter.ai/api/v1"
-key_required = true
-
-[[models]]
-id = "claude-opus-latest"
-display_name = "Anthropic: Claude Opus Latest"
-tier = "frontier"
-context_window = 1000000
-max_output_tokens = 128000
-input_cost_per_m = 5.0
-output_cost_per_m = 25.0
-supports_streaming = true