feat(gateway): add config mounted

docker-compose.yml

@@ -45,7 +45,7 @@ services:
       - "0.0.0.0:4000:3000"
     volumes:
       - llm-gateway-data:/data
-      - ./llm-gateway/configs/config.yaml:/etc/llm-gateway/config.yaml:ro
+      - ./llm-gateway.yaml:/etc/llm-gateway/config.yaml:ro
     environment:
       - SESSION_SECRET=${SESSION_SECRET}
       - ADMIN_USERNAME=${ADMIN_USERNAME}

llm-gateway.yaml

@@ -0,0 +1,264 @@
+server:
+  listen: "0.0.0.0:3000"
+  request_timeout: 300s
+  max_request_body_mb: 10
+  session_secret: "${SESSION_SECRET}"
+  default_admin:
+    username: "${ADMIN_USERNAME}"
+    password: "${ADMIN_PASSWORD}"
+
+tokens:
+  - name: "open-webui"
+    key: "${OPENWEBUI_API_KEY}"
+    rate_limit_rpm: 0          # unlimited
+    daily_budget_usd: 5.0
+  - name: "rayandrew"
+    key: "${PERSONAL_API_KEY}"
+    rate_limit_rpm: 0          # unlimited
+    daily_budget_usd: 10.0
+
+pricing_lookup:
+  # url: "https://raw.githubusercontent.com/pydantic/genai-prices/main/prices/data_slim.json"  # default
+  refresh_interval: 6h
+
+database:
+  path: "/data/gateway.db"
+  retention_days: 90
+
+cache:
+  enabled: true
+  address: "valkey:6379"
+  ttl: 3600
+
+providers:
+  - name: deepinfra
+    base_url: "https://api.deepinfra.com/v1/openai"
+    api_key: "${DEEPINFRA_API_KEY}"
+    priority: 1
+    timeout: 120s
+  - name: siliconflow
+    base_url: "https://api.siliconflow.com/v1"
+    api_key: "${SILICONFLOW_API_KEY}"
+    priority: 2
+    timeout: 120s
+  - name: openrouter
+    base_url: "https://openrouter.ai/api/v1"
+    api_key: "${OPENROUTER_API_KEY}"
+    priority: 3
+    timeout: 120s
+  - name: groq
+    base_url: "https://api.groq.com/openai/v1"
+    api_key: "${GROQ_API_KEY}"
+    priority: 1
+    timeout: 120s
+  - name: cerebras
+    base_url: "https://api.cerebras.ai/v1"
+    api_key: "${CEREBRAS_API_KEY}"
+    priority: 1
+    timeout: 120s
+
+models:
+  # ── DeepSeek V3.2 ──
+  - name: "deepseek-v3.2"
+    routes:
+      - provider: deepinfra
+        model: "deepseek-ai/DeepSeek-V3.2"
+        pricing: { input: 0.26, output: 0.38 }
+      - provider: siliconflow
+        model: "deepseek-ai/DeepSeek-V3.2"
+        pricing: { input: 0.27, output: 0.42 }
+      - provider: openrouter
+        model: "deepseek/deepseek-chat-v3-0324"
+        pricing: { input: 0.30, output: 0.88 }
+
+  # ── DeepSeek R1 ──
+  - name: "deepseek-r1"
+    routes:
+      - provider: deepinfra
+        model: "deepseek-ai/DeepSeek-R1"
+        pricing: { input: 0.40, output: 1.60 }
+      - provider: openrouter
+        model: "deepseek/deepseek-r1"
+        pricing: { input: 0.55, output: 2.19 }
+
+  # ── GPT-OSS (OpenAI open-weight MoE) ──
+  - name: "gpt-oss"
+    routes:
+      - provider: deepinfra
+        model: "openai/gpt-oss-120b"
+        pricing: { input: 0.05, output: 0.24 }
+
+  - name: "gpt-oss-20b"
+    routes:
+      - provider: deepinfra
+        model: "openai/gpt-oss-20b"
+        pricing: { input: 0.04, output: 0.16 }
+
+  # ── Nemotron ──
+  - name: "nemotron-super"
+    routes:
+      - provider: deepinfra
+        model: "nvidia/Llama-3.3-Nemotron-Super-49B-v1.5"
+        pricing: { input: 0.10, output: 0.40 }
+
+  - name: "nemotron-nano"
+    routes:
+      - provider: deepinfra
+        model: "nvidia/NVIDIA-Nemotron-Nano-9B-v2"
+        pricing: { input: 0.04, output: 0.16 }
+
+  # ── Devstral ──
+  - name: "devstral"
+    routes:
+      - provider: deepinfra
+        model: "mistralai/Devstral-Small-2505"
+
+  # ── GLM ──
+  - name: "glm-4.6"
+    routes:
+      - provider: deepinfra
+        model: "zai-org/GLM-4.6"
+        pricing: { input: 0.60, output: 1.90 }
+
+  - name: "glm-4.7"
+    routes:
+      - provider: deepinfra
+        model: "zai-org/GLM-4.7"
+        pricing: { input: 0.40, output: 1.75 }
+      - provider: siliconflow
+        model: "THUDM/GLM-4-32B-0414"
+
+  - name: "glm-5"
+    routes:
+      - provider: deepinfra
+        model: "zai-org/GLM-5"
+        pricing: { input: 0.80, output: 2.56 }
+
+  # ── Kimi ──
+  - name: "kimi-k2"
+    routes:
+      - provider: deepinfra
+        model: "moonshotai/Kimi-K2-Instruct-0905"
+        pricing: { input: 0.50, output: 2.00 }
+      - provider: siliconflow
+        model: "moonshotai/Kimi-K2-Instruct-0905"
+        pricing: { input: 0.58, output: 2.29 }
+
+  - name: "kimi-k2.5"
+    routes:
+      - provider: deepinfra
+        model: "moonshotai/Kimi-K2.5"
+        pricing: { input: 0.45, output: 2.25 }
+      - provider: openrouter
+        model: "moonshotai/kimi-k2.5"
+
+  # ── Qwen3 Coder ──
+  - name: "qwen3-coder"
+    routes:
+      - provider: siliconflow
+        model: "Qwen/Qwen3-Coder-480B-A35B-Instruct"
+        pricing: { input: 1.14, output: 2.28 }
+
+  - name: "qwen3-coder-30b"
+    routes:
+      - provider: siliconflow
+        model: "Qwen/Qwen3-Coder-30B-A3B-Instruct"
+
+  # ── Llama ──
+  - name: "llama-3.3-70b"
+    routes:
+      - provider: groq
+        model: "llama-3.3-70b-versatile"
+      - provider: cerebras
+        model: "llama-3.3-70b"
+      - provider: deepinfra
+        model: "meta-llama/Llama-3.3-70B-Instruct"
+        pricing: { input: 0.23, output: 0.40 }
+
+  - name: "llama-3.1-8b"
+    routes:
+      - provider: groq
+        model: "llama-3.1-8b-instant"
+      - provider: cerebras
+        model: "llama-3.1-8b"
+      - provider: deepinfra
+        model: "meta-llama/Meta-Llama-3.1-8B-Instruct"
+        pricing: { input: 0.03, output: 0.05 }
+
+  # ── Qwen 2.5 ──
+  - name: "qwen-2.5-72b"
+    routes:
+      - provider: groq
+        model: "qwen-2.5-72b"
+      - provider: deepinfra
+        model: "Qwen/Qwen2.5-72B-Instruct"
+        pricing: { input: 0.23, output: 0.40 }
+
+  - name: "qwen-2.5-coder-32b"
+    routes:
+      - provider: groq
+        model: "qwen-2.5-coder-32b"
+      - provider: deepinfra
+        model: "Qwen/Qwen2.5-Coder-32B-Instruct"
+        pricing: { input: 0.07, output: 0.16 }
+
+  # ── Other ──
+  - name: "gemma-2-9b"
+    routes:
+      - provider: groq
+        model: "gemma2-9b-it"
+
+  - name: "deepseek-r1-distill-llama-70b"
+    routes:
+      - provider: groq
+        model: "deepseek-r1-distill-llama-70b"
+      - provider: deepinfra
+        model: "deepseek-ai/DeepSeek-R1-Distill-Llama-70B"
+        pricing: { input: 0.23, output: 0.69 }
+
+  - name: "deepseek-r1-distill-qwen-32b"
+    routes:
+      - provider: deepinfra
+        model: "deepseek-ai/DeepSeek-R1-Distill-Qwen-32B"
+        pricing: { input: 0.07, output: 0.16 }
+
+  # ── OpenRouter premium ──
+  - name: "deepseek-v3-free"
+    routes:
+      - provider: openrouter
+        model: "deepseek/deepseek-chat-v3-0324:free"
+
+  - name: "minimax-m2.5"
+    routes:
+      - provider: openrouter
+        model: "minimax/minimax-m2.5"
+
+  - name: "gpt-4.1-mini"
+    routes:
+      - provider: openrouter
+        model: "openai/gpt-4.1-mini"
+
+  - name: "gpt-4.1"
+    routes:
+      - provider: openrouter
+        model: "openai/gpt-4.1"
+
+  - name: "gemini-3-flash-preview"
+    routes:
+      - provider: openrouter
+        model: "google/gemini-3-flash-preview"
+
+  - name: "gemini-2.5-pro"
+    routes:
+      - provider: openrouter
+        model: "google/gemini-2.5-pro-preview"
+
+  - name: "claude-sonnet"
+    routes:
+      - provider: openrouter
+        model: "anthropic/claude-sonnet-4"
+
+  - name: "trinity-large-preview"
+    routes:
+      - provider: openrouter
+        model: "arcee-ai/trinity-large-preview"