1 line
10 KiB
JSON
1 line
10 KiB
JSON
{"repo": "bejugamvarun/analytics-agent", "files": {"src/risk_analytics_agent/config/models.yaml": {"732853cee5dcea316eeea92a60e4ce0d174401d9": {"t": 1785746323.543965, "n": 3839, "h": "8d4097db39a7eb4f", "c": "# LLM model configuration\n# Supports on-prem hosted models via LiteLLM-compatible endpoints.\n# Set `provider` to match your deployment: \"openai\", \"azure\", \"anthropic\",\n# \"vllm\", \"ollama\", or any provider LiteLLM supports.\n#\n# `default` applies to every agent unless overridden under `agents`.\n# Per-agent entries inherit all default values — only specify the fields\n# you want to change (e.g. a different model_name or temperature).\n\ndefault:\n provider: openai # LM Studio uses OpenAI-compatible API\n model_name: openai/gpt-oss-20b # Use the model name from LM Studio (can be anything)\n api_base: http://localhost:1234/v1 # LM Studio default endpoint\n api_key: lm-studio # Placeholder (LM Studio doesn't need a real key)\n temperature: 0.7 # Temperature for non-reasoning models\n max_tokens: 8192 # Max output tokens\n # timeout: 120\n \n # NOTE: Reasoning model parameters (reasoning_effort, supports_reasoning, etc.)\n # should ONLY be added to specific agents that use reasoning models (o1, DeepSeek R1)\n # Don't add them to default config unless all your models support reasoning!\n \n # Alternative providers (uncomment to use):\n # provider: anthropic\n # model_name: claude-haiku-4-5\n # api_base: ${LLM_API_BASE}\n # api_key: ${LLM_API_KEY}\n\n# Per-agent overrides (uncomment / add as needed).\n# Any key not specified here falls back to the default above.\nagents:\n # ============================================================================\n # STANDARD MODEL EXAMPLE\n # ============================================================================\n # orchestrator_agent:\n # model_name: claude-opus-4-6\n # temperature: 0.7\n # max_tokens: 16384\n \n # ============================================================================\n # REASONING MODEL EXAMPLES (o1, DeepSeek R1, etc.)\n # IMPORTANT: Only use reasoning parameters for actual reasoning models!\n # ============================================================================\n \n # Example 1: OpenAI o1-preview (Best reasoning, highest cost)\n # quantitative_analysis_agent:\n # provider: openai\n # model_name: o1-preview\n # reasoning_effort: high # ONLY for o1 models\n # supports_reasoning: true # Flags this as reasoning model\n # max_completion_tokens: 32768 # o1 uses this instead of max_tokens\n # api_key: ${OPENAI_API_KEY}\n # # NOTE: o1 models don't support temperature/top_p - auto-removed when supports_reasoning=true\n\n # Example 2: OpenAI o1-mini (Fast reasoning, medium cost)\n # variance_analysis_agent:\n # provider: openai\n # model_name: o1-mini\n # reasoning_effort: medium\n # supports_reasoning: true\n # max_completion_tokens: 16384\n # api_key: ${OPENAI_API_KEY}\n\n # Example 3: DeepSeek R1 (Cost-effective reasoning)\n # anomaly_detection_agent:\n # provider: openai\n # model_name: deepseek-reasoner\n # api_base: https://api.deepseek.com/v1\n # api_key: ${DEEPSEEK_API_KEY}\n # supports_reasoning: true # Enables reasoning-specific handling\n # max_tokens: 8192\n # temperature: 0.7 # DeepSeek supports temperature\n \n # ============================================================================\n # OTHER EXAMPLES\n # ============================================================================\n \n # Fast model for simple tasks\n # schema_discovery_agent:\n # provider: openai\n # model_name: gpt-4o-mini\n # temperature: 0.5\n # max_tokens: 4096\n\n # Code-tuned model for SQL\n # data_retrieval_agent:\n # model_name: codestral-latest\n # temperature: 0.3\n\n # Local vLLM instance\n # report_generation_agent:\n # provider: vllm\n # model_name: mistral-7b\n # api_base: https://vllm.internal.corp/v1\n # api_key: your-key\n"}, "cf8b796ec54ab725ad7b461c28e4a5a0f96621cf": {"t": 1785761830.3368397, "n": 3839, "h": "8d4097db39a7eb4f", "c": "# LLM model configuration\n# Supports on-prem hosted models via LiteLLM-compatible endpoints.\n# Set `provider` to match your deployment: \"openai\", \"azure\", \"anthropic\",\n# \"vllm\", \"ollama\", or any provider LiteLLM supports.\n#\n# `default` applies to every agent unless overridden under `agents`.\n# Per-agent entries inherit all default values — only specify the fields\n# you want to change (e.g. a different model_name or temperature).\n\ndefault:\n provider: openai # LM Studio uses OpenAI-compatible API\n model_name: openai/gpt-oss-20b # Use the model name from LM Studio (can be anything)\n api_base: http://localhost:1234/v1 # LM Studio default endpoint\n api_key: lm-studio # Placeholder (LM Studio doesn't need a real key)\n temperature: 0.7 # Temperature for non-reasoning models\n max_tokens: 8192 # Max output tokens\n # timeout: 120\n \n # NOTE: Reasoning model parameters (reasoning_effort, supports_reasoning, etc.)\n # should ONLY be added to specific agents that use reasoning models (o1, DeepSeek R1)\n # Don't add them to default config unless all your models support reasoning!\n \n # Alternative providers (uncomment to use):\n # provider: anthropic\n # model_name: claude-haiku-4-5\n # api_base: ${LLM_API_BASE}\n # api_key: ${LLM_API_KEY}\n\n# Per-agent overrides (uncomment / add as needed).\n# Any key not specified here falls back to the default above.\nagents:\n # ============================================================================\n # STANDARD MODEL EXAMPLE\n # ============================================================================\n # orchestrator_agent:\n # model_name: claude-opus-4-6\n # temperature: 0.7\n # max_tokens: 16384\n \n # ============================================================================\n # REASONING MODEL EXAMPLES (o1, DeepSeek R1, etc.)\n # IMPORTANT: Only use reasoning parameters for actual reasoning models!\n # ============================================================================\n \n # Example 1: OpenAI o1-preview (Best reasoning, highest cost)\n # quantitative_analysis_agent:\n # provider: openai\n # model_name: o1-preview\n # reasoning_effort: high # ONLY for o1 models\n # supports_reasoning: true # Flags this as reasoning model\n # max_completion_tokens: 32768 # o1 uses this instead of max_tokens\n # api_key: ${OPENAI_API_KEY}\n # # NOTE: o1 models don't support temperature/top_p - auto-removed when supports_reasoning=true\n\n # Example 2: OpenAI o1-mini (Fast reasoning, medium cost)\n # variance_analysis_agent:\n # provider: openai\n # model_name: o1-mini\n # reasoning_effort: medium\n # supports_reasoning: true\n # max_completion_tokens: 16384\n # api_key: ${OPENAI_API_KEY}\n\n # Example 3: DeepSeek R1 (Cost-effective reasoning)\n # anomaly_detection_agent:\n # provider: openai\n # model_name: deepseek-reasoner\n # api_base: https://api.deepseek.com/v1\n # api_key: ${DEEPSEEK_API_KEY}\n # supports_reasoning: true # Enables reasoning-specific handling\n # max_tokens: 8192\n # temperature: 0.7 # DeepSeek supports temperature\n \n # ============================================================================\n # OTHER EXAMPLES\n # ============================================================================\n \n # Fast model for simple tasks\n # schema_discovery_agent:\n # provider: openai\n # model_name: gpt-4o-mini\n # temperature: 0.5\n # max_tokens: 4096\n\n # Code-tuned model for SQL\n # data_retrieval_agent:\n # model_name: codestral-latest\n # temperature: 0.3\n\n # Local vLLM instance\n # report_generation_agent:\n # provider: vllm\n # model_name: mistral-7b\n # api_base: https://vllm.internal.corp/v1\n # api_key: your-key\n"}, "5ae6161d64424c6a605c4f091cb59610b508cd1c": {"t": 1785761834.3609178, "n": 1810, "h": "58086e82590829f2", "c": "# LLM model configuration\n# Supports on-prem hosted models via LiteLLM-compatible endpoints.\n# Set `provider` to match your deployment: \"openai\", \"azure\", \"anthropic\",\n# \"vllm\", \"ollama\", or any provider LiteLLM supports.\n#\n# `default` applies to every agent unless overridden under `agents`.\n# Per-agent entries inherit all default values — only specify the fields\n# you want to change (e.g. a different model_name or temperature).\n\ndefault:\n provider: openai # LM Studio uses OpenAI-compatible API\n model_name: unsloth/gpt-oss-20b # Use the model name from LM Studio (can be anything)\n api_base: http://localhost:1234/v1 # LM Studio default endpoint\n api_key: lm-studio # Placeholder (LM Studio doesn't need a real key)\n # provider: anthropic # LiteLLM provider prefix\n # model_name: claude-haiku-4-5\n # api_base: ${LLM_API_BASE} # Your on-prem endpoint URL\n # api_key: ${LLM_API_KEY} # API key for the endpoint\n # temperature: 0.0\n # max_tokens: 8192\n # timeout: 120\n # custom_llm_provider: anthropic\n\n# Per-agent overrides (uncomment / add as needed).\n# Any key not specified here falls back to the default above.\nagents:\n # Example: use a larger model for the orchestrator\n # orchestrator_agent:\n # model_name: claude-opus-4-6\n # max_tokens: 16384\n\n # Example: use a fast, cheap model for schema discovery\n # schema_discovery_agent:\n # provider: vllm\n # model_name: mistral-7b\n # api_base: https://vllm.internal.corp/v1\n\n # Example: use a code-tuned model for data retrieval SQL generation\n # data_retrieval_agent:\n # model_name: codestral-latest\n\n # quantitative_analysis_agent:\n # variance_analysis_agent:\n # drilldown_analysis_agent:\n # anomaly_detection_agent:\n # report_generation_agent:\n"}}}} |