Worker Tools

delegate_task

Route a task to a cheaper model, or summarize vault files automatically.

# Task delegation
delegate_task(
    prompt="Explain this regex: ^(?:[a-z0-9]+\.)*[a-z0-9]+$",
    context="",              # optional context to include
    max_cost_per_request=0,  # 0 = free models only
    model=""                 # explicit model override
)

# Vault file summarization
delegate_task(
    project="my-project",
    section="context",       # or use path="..."
    max_summary_lines=20     # target summary length
)

Task Delegation

When prompt is provided, tasks are routed through tiers in order:

1. Ollama (local) — Free. Best for trivial tasks. Falls through if unavailable.
2. OpenRouter free — Free tier models (e.g., Qwen3 Coder 480B). Real code work.
3. OpenRouter paid — Only when max_cost_per_request > 0 and monthly budget allows. Model configurable via HIVE_OPENROUTER_PAID_MODEL.
4. Reject — Returns error so the host handles the task directly.

Vault Summarization

When project is provided, reads a vault file:

• Small files (≤50 lines): returned directly with metadata
• Large files (>50 lines): auto-delegated to a worker for summarization. Falls back to raw content if workers are unavailable.

Explicit Model Selection

Skip routing and target a specific provider:

# Force local Ollama
delegate_task(prompt="...", model="ollama:qwen2.5-coder:7b")

# Force OpenRouter free
delegate_task(prompt="...", model="openrouter:qwen/qwen3-coder:free")

# Force OpenRouter paid
delegate_task(prompt="...", model="openrouter:qwen/qwen3-coder", max_cost_per_request=0.01)

Response Format

Each response includes a metadata footer:

[model: qwen2.5-coder:7b | provider: ollama | cost: $0.00 | latency: 2.1s]

worker_status

Show worker health, budget, and available models.

# Full status with model listing
worker_status()

# Status without model listing
worker_status(include_models=False)

Returns:

• Monthly budget remaining (spent / cap)
• Ollama connectivity status
• OpenRouter connectivity status
• Request counts and cost breakdown
• Available models across all providers (when include_models=True)