Models

from agentfield import Agent, AIConfig, HarnessConfig

# Agent-level config — defaults for all .ai() calls
app = Agent(
    node_id="production-agent",
    ai_config=AIConfig(
        model="anthropic/claude-sonnet-4-20250514",       # default model
        fallback_models=[                                  # auto-failover chain
            "openai/gpt-4o",
            "deepseek/deepseek-chat",
        ],
        max_cost_per_call=0.05,                            # hard cap per call
        daily_budget=10.00,                                # daily spend limit
        enable_rate_limit_retry=True,                      # auto-retry with backoff
        rate_limit_max_retries=10,
        auto_inject_memory=["user_prefs", "conversation"], # inject memory into prompts
    ),
    # Harness uses a DIFFERENT model config — coding agents have their own budget
    harness_config=HarnessConfig(
        provider="claude-code",
        model="sonnet",
        max_budget_usd=2.00,       # separate cost cap for coding tasks
        max_turns=30,
    ),
)

# Per-call model override — no agent reconfiguration needed
category = await app.ai(
    user=ticket_text,
    schema=TicketCategory,
    model="openai/gpt-4o-mini",   # cheap model for fast classification
)

analysis = await app.ai(
    user=ticket_text,
    schema=DeepAnalysis,
    model="anthropic/claude-sonnet-4-20250514",  # powerful model for reasoning
    temperature=0.0,                              # deterministic
)

# Local models — zero data leaves your machine
local_app = Agent(
    node_id="local-agent",
    ai_config=AIConfig(
        model="ollama/llama3",
        api_base="http://localhost:11434",
    ),
)

import { Agent } from '@agentfield/sdk';

const agent = new Agent({
  nodeId: 'production-agent',
  aiConfig: {
    provider: 'anthropic',
    model: 'claude-sonnet-4-20250514',             // default model
    temperature: 0.3,
    maxTokens: 4096,
    enableRateLimitRetry: true,                    // auto-retry with backoff
    rateLimitMaxRetries: 20,
  },
});

agent.reasoner('analyze', async (ctx) => {
  // Per-call model override — cheap for routing, powerful for analysis
  const category = await ctx.ai(ticketText, {
    model: 'gpt-4o-mini',                         // fast + cheap
    schema: TicketCategory,
  });

  const analysis = await ctx.ai(ticketText, {
    model: 'claude-sonnet-4-20250514',             // deep reasoning
    schema: DeepAnalysis,
    temperature: 0.0,
  });

  // Harness config — different model for coding tasks
  const fix = await agent.harness('Fix the failing test in auth.test.ts', {
    provider: 'codex',
    model: 'o4-mini',
    maxBudgetUsd: 1.00,
  });

  return { category, analysis };
});

// Agent-level config
config := &ai.Config{
    APIKey:  os.Getenv("OPENROUTER_API_KEY"),
    BaseURL: "https://openrouter.ai/api/v1",
    Model:   "anthropic/claude-sonnet-4-20250514",  // default model
}
client, _ := ai.NewClient(config)

// Per-call model override
catResp, _ := client.Complete(ctx, ticketText,
    ai.WithModel("gpt-4o-mini"),                    // cheap for classification
    ai.WithSchema(TicketCategory{}),
)

analysisResp, _ := client.Complete(ctx, ticketText,
    ai.WithModel("anthropic/claude-sonnet-4-20250514"), // powerful for analysis
    ai.WithTemperature(0.0),
    ai.WithSchema(DeepAnalysis{}),
)