Add config.py — model registry, prompts, tools

multeclaw/config.py

@@ -0,0 +1,272 @@
+"""
+Multeclaw Configuration — model registry, defaults, and system constants.
+"""
+
+from dataclasses import dataclass, field
+from enum import Enum
+from typing import Optional
+import os
+
+
+# ─── Model Provider Enum ──────────────────────────────────────────────────────
+class Provider(str, Enum):
+    OPENAI = "openai"
+    ANTHROPIC = "anthropic"
+    HUGGINGFACE = "huggingface"
+    OLLAMA = "ollama"
+    GROQ = "groq"
+
+
+# ─── Model Definition ─────────────────────────────────────────────────────────
+@dataclass
+class ModelDef:
+    display_name: str
+    provider: Provider
+    model_id: str
+    description: str = ""
+    max_tokens: int = 4096
+    supports_streaming: bool = True
+    supports_tools: bool = True
+    context_window: int = 128000
+    tier: str = "standard"  # "fast", "standard", "strong"
+
+
+# ─── Model Registry ───────────────────────────────────────────────────────────
+MODEL_REGISTRY: dict[str, ModelDef] = {
+    # OpenAI
+    "GPT-4o": ModelDef(
+        display_name="GPT-4o",
+        provider=Provider.OPENAI,
+        model_id="gpt-4o",
+        description="Most capable OpenAI model — strong reasoning + vision",
+        max_tokens=4096,
+        context_window=128000,
+        tier="strong",
+    ),
+    "GPT-4o Mini": ModelDef(
+        display_name="GPT-4o Mini",
+        provider=Provider.OPENAI,
+        model_id="gpt-4o-mini",
+        description="Fast, cost-efficient OpenAI model",
+        max_tokens=4096,
+        context_window=128000,
+        tier="fast",
+    ),
+    "GPT-4 Turbo": ModelDef(
+        display_name="GPT-4 Turbo",
+        provider=Provider.OPENAI,
+        model_id="gpt-4-turbo",
+        description="Previous-gen GPT-4 with vision",
+        max_tokens=4096,
+        context_window=128000,
+        tier="standard",
+    ),
+    # Anthropic
+    "Claude 4 Opus": ModelDef(
+        display_name="Claude 4 Opus",
+        provider=Provider.ANTHROPIC,
+        model_id="claude-opus-4-5",
+        description="Most capable Claude model — deep reasoning",
+        max_tokens=4096,
+        context_window=200000,
+        tier="strong",
+    ),
+    "Claude 4 Sonnet": ModelDef(
+        display_name="Claude 4 Sonnet",
+        provider=Provider.ANTHROPIC,
+        model_id="claude-sonnet-4-5",
+        description="Balanced Claude model — strong + fast",
+        max_tokens=4096,
+        context_window=200000,
+        tier="standard",
+    ),
+    "Claude 3.5 Haiku": ModelDef(
+        display_name="Claude 3.5 Haiku",
+        provider=Provider.ANTHROPIC,
+        model_id="claude-3-5-haiku-20241022",
+        description="Fastest Claude model",
+        max_tokens=4096,
+        context_window=200000,
+        tier="fast",
+    ),
+    # HuggingFace (via inference providers)
+    "Llama 3 8B": ModelDef(
+        display_name="Llama 3 8B",
+        provider=Provider.HUGGINGFACE,
+        model_id="meta-llama/Meta-Llama-3-8B-Instruct",
+        description="Meta open-source 8B model via HF Inference",
+        max_tokens=2048,
+        context_window=8192,
+        tier="fast",
+        supports_tools=False,
+    ),
+    "Qwen 2.5 72B": ModelDef(
+        display_name="Qwen 2.5 72B",
+        provider=Provider.HUGGINGFACE,
+        model_id="Qwen/Qwen2.5-72B-Instruct",
+        description="Qwen large model via HF Inference",
+        max_tokens=4096,
+        context_window=32768,
+        tier="strong",
+        supports_tools=False,
+    ),
+    "Mistral 7B": ModelDef(
+        display_name="Mistral 7B",
+        provider=Provider.HUGGINGFACE,
+        model_id="mistralai/Mistral-7B-Instruct-v0.3",
+        description="Mistral efficient 7B via HF Inference",
+        max_tokens=2048,
+        context_window=32768,
+        tier="fast",
+        supports_tools=False,
+    ),
+    # Groq (fast inference)
+    "Llama 3 70B (Groq)": ModelDef(
+        display_name="Llama 3 70B (Groq)",
+        provider=Provider.GROQ,
+        model_id="llama3-70b-8192",
+        description="Llama 3 70B on Groq — extremely fast inference",
+        max_tokens=4096,
+        context_window=8192,
+        tier="standard",
+        supports_tools=True,
+    ),
+    "Mixtral 8x7B (Groq)": ModelDef(
+        display_name="Mixtral 8x7B (Groq)",
+        provider=Provider.GROQ,
+        model_id="mixtral-8x7b-32768",
+        description="Mixtral MoE on Groq",
+        max_tokens=4096,
+        context_window=32768,
+        tier="fast",
+        supports_tools=False,
+    ),
+    # Ollama (local)
+    "Ollama: Llama 3": ModelDef(
+        display_name="Ollama: Llama 3",
+        provider=Provider.OLLAMA,
+        model_id="llama3",
+        description="Local Llama 3 via Ollama",
+        max_tokens=4096,
+        context_window=8192,
+        tier="standard",
+        supports_tools=False,
+    ),
+    "Ollama: Mistral": ModelDef(
+        display_name="Ollama: Mistral",
+        provider=Provider.OLLAMA,
+        model_id="mistral",
+        description="Local Mistral via Ollama",
+        max_tokens=4096,
+        context_window=32768,
+        tier="fast",
+        supports_tools=False,
+    ),
+}
+
+
+# ─── Agent System Prompts ─────────────────────────────────────────────────────
+SYSTEM_PROMPTS = {
+    "Multeclaw Agent": (
+        "You are Multeclaw, a high-capability general-purpose AI agent system. "
+        "You interpret user intent, decompose complex objectives, reason step by step, "
+        "and deliver precise, verified results. You are calm, direct, and thorough. "
+        "You never invent facts. You state limitations clearly. "
+        "For complex tasks, you think step by step before answering."
+    ),
+    "Code Expert": (
+        "You are a senior software engineer. Write clean, well-documented, production-grade code. "
+        "Follow best practices. Explain your design choices briefly. Include error handling."
+    ),
+    "Research Analyst": (
+        "You are a thorough research analyst. Provide accurate, well-structured analysis "
+        "with clear citations of reasoning. Distinguish facts from inferences. "
+        "Present findings in organized sections with key takeaways."
+    ),
+    "Creative Writer": (
+        "You are a creative writer with a vivid, engaging style. "
+        "Write with strong imagery, varied sentence structure, and emotional depth. "
+        "Adapt your tone to the request — formal, casual, poetic, or technical."
+    ),
+    "Data Scientist": (
+        "You are a data science expert. Provide rigorous statistical analysis, "
+        "suggest appropriate methods, explain trade-offs, and write clean analysis code. "
+        "Always consider data quality, biases, and limitations."
+    ),
+    "Custom": "",
+}
+
+
+# ─── Agent Loop Config ─────────────────────────────────────────────────────────
+@dataclass
+class AgentConfig:
+    max_reasoning_steps: int = 10
+    max_tool_retries: int = 3
+    max_repair_attempts: int = 2
+    verification_enabled: bool = True
+    streaming_enabled: bool = True
+    temperature: float = 0.7
+    top_p: float = 1.0
+    presence_penalty: float = 0.0
+    frequency_penalty: float = 0.0
+
+
+# ─── Built-in Tools ────────────────────────────────────────────────────────────
+BUILT_IN_TOOLS = {
+    "web_search": {
+        "name": "web_search",
+        "description": "Search the web for current information. Use when you need up-to-date facts.",
+        "parameters": {
+            "type": "object",
+            "properties": {
+                "query": {"type": "string", "description": "Search query"}
+            },
+            "required": ["query"],
+        },
+    },
+    "calculator": {
+        "name": "calculator",
+        "description": "Evaluate a mathematical expression. Supports Python math syntax.",
+        "parameters": {
+            "type": "object",
+            "properties": {
+                "expression": {"type": "string", "description": "Math expression like '2**10 + sqrt(144)'"}
+            },
+            "required": ["expression"],
+        },
+    },
+    "code_executor": {
+        "name": "code_executor",
+        "description": "Execute Python code in a sandboxed environment. Returns stdout/stderr.",
+        "parameters": {
+            "type": "object",
+            "properties": {
+                "code": {"type": "string", "description": "Python code to execute"}
+            },
+            "required": ["code"],
+        },
+    },
+    "file_reader": {
+        "name": "file_reader",
+        "description": "Read the contents of a text file.",
+        "parameters": {
+            "type": "object",
+            "properties": {
+                "path": {"type": "string", "description": "File path to read"}
+            },
+            "required": ["path"],
+        },
+    },
+    "file_writer": {
+        "name": "file_writer",
+        "description": "Write content to a file. Creates parent directories if needed.",
+        "parameters": {
+            "type": "object",
+            "properties": {
+                "path": {"type": "string", "description": "File path to write"},
+                "content": {"type": "string", "description": "Content to write"},
+            },
+            "required": ["path", "content"],
+        },
+    },
+}