swarm（swarm 搜索）

Swarm — Cut Your LLM Costs by 200x

Turn your expensive model into an affordable daily driver. Offload the boring stuff to Gemini Flash workers — parallel, batch, research — at a fraction of the cost.

At a Glance

When to Use

Swarm is ideal for:

• 3+ independent tasks (research, summaries, comparisons)
• Comparing or researching multiple subjects
• Multiple URLs to fetch/analyze
• Batch processing (documents, entities, facts)
• Complex analysis needing multiple perspectives → use chain

Quick Reference

# Check daemon (do this every session)
swarm status

# Start if not running
swarm start

# Parallel prompts
swarm parallel "What is X?" "What is Y?" "What is Z?"

# Research multiple subjects
swarm research "OpenAI" "Anthropic" "Mistral" --topic "AI safety"

# Discover capabilities
swarm capabilities

Execution Modes

Parallel (v1.0)

N prompts → N workers simultaneously. Best for independent tasks.

swarm parallel "prompt1" "prompt2" "prompt3"

Research (v1.1)

Multi-phase: search → fetch → analyze. Uses Google Search grounding.

swarm research "Buildertrend" "Jobber" --topic "pricing 2026"

Chain (v1.3) — Refinement Pipelines

Data flows through multiple stages, each with a different perspective/filter. Stages run in sequence; tasks within a stage run in parallel.

Stage modes:

• parallel — N inputs → N workers (same perspective)
• single — merged input → 1 worker
• fan-out — 1 input → N workers with DIFFERENT perspectives
• reduce — N inputs → 1 synthesized output

Auto-chain — describe what you want, get an optimal pipeline:

curl -X POST http://localhost:9999/chain/auto \
  -d '{"task":"Find business opportunities","data":"...market data...","depth":"standard"}'

Manual chain:

swarm chain pipeline.json
# or
echo '{"stages":[...]}' | swarm chain --stdin

Depth presets: quick (2 stages), standard (4), deep (6), exhaustive (8)

Built-in perspectives: extractor, filter, enricher, analyst, synthesizer, challenger, optimizer, strategist, researcher, critic

Preview without executing:

curl -X POST http://localhost:9999/chain/preview \
  -d '{"task":"...","depth":"standard"}'

Benchmark (v1.3)

Compare single vs parallel vs chain on the same task with LLM-as-judge scoring.

curl -X POST http://localhost:9999/benchmark \
  -d '{"task":"Analyze X","data":"...","depth":"standard"}'

Scores on 6 FLASK dimensions: accuracy (2x weight), depth (1.5x), completeness, coherence, actionability (1.5x), nuance.

Capabilities Discovery (v1.3)

Lets the orchestrator discover what execution modes are available:

swarm capabilities
# or
curl http://localhost:9999/capabilities

Prompt Cache (v1.3.2)

LRU cache for LLM responses. 212x speedup on cache hits (parallel), 514x on chains.

• Keyed by hash of instruction + input + perspective
• 500 entries max, 1 hour TTL
• Skips web search tasks (need fresh data)
• Persists to disk across daemon restarts
• Per-task bypass: set task.cache = false

# View cache stats
curl http://localhost:9999/cache

# Clear cache
curl -X DELETE http://localhost:9999/cache

Cache stats show in swarm status.

Stage Retry (v1.3.2)

If tasks fail within a chain stage, only the failed tasks get retried (not the whole stage). Default: 1 retry. Configurable per-phase via phase.retries or globally via options.stageRetries.

Cost Tracking (v1.3.1)

All endpoints return cost data in their complete event:

• session — current daemon session totals
• daily — persisted across restarts, accumulates all day

swarm status        # Shows session + daily cost
swarm savings       # Monthly savings report

Web Search (v1.1)

Workers search the live web via Google Search grounding (Gemini only, no extra cost).

# Research uses web search by default
swarm research "Subject" --topic "angle"

# Parallel with web search
curl -X POST http://localhost:9999/parallel \
  -d '{"prompts":["Current price of X?"],"options":{"webSearch":true}}'

JavaScript API

const { parallel, research } = require('~/clawd/skills/node-scaling/lib');
const { SwarmClient } = require('~/clawd/skills/node-scaling/lib/client');

// Simple parallel
const result = await parallel(['prompt1', 'prompt2', 'prompt3']);

// Client with streaming
const client = new SwarmClient();
for await (const event of client.parallel(prompts)) { ... }
for await (const event of client.research(subjects, topic)) { ... }

// Chain
const result = await client.chainSync({ task, data, depth });

Daemon Management

swarm start              # Start daemon (background)
swarm stop               # Stop daemon
swarm status             # Status, cost, cache stats
swarm restart            # Restart daemon
swarm savings            # Monthly savings report
swarm logs [N]           # Last N lines of daemon log

Performance (v1.3.2)

Config

Location: ~/.config/clawdbot/node-scaling.yaml

node_scaling:
  enabled: true
  limits:
    max_nodes: 16
    max_concurrent_api: 16
  provider:
    name: gemini
    model: gemini-2.0-flash
  web_search:
    enabled: true
    parallel_default: false
  cost:
    max_daily_spend: 10.00

Troubleshooting

Structured Output (v1.3.7)

Force JSON output with schema validation — zero parse failures on structured tasks.

# With built-in schema
curl -X POST http://localhost:9999/structured \
  -d '{"prompt":"Extract entities from: Tim Cook announced iPhone 17","schema":"entities"}'

# With custom schema
curl -X POST http://localhost:9999/structured \
  -d '{"prompt":"Classify this text","data":"...","schema":{"type":"object","properties":{"category":{"type":"string"}}}}'

# JSON mode (no schema, just force JSON)
curl -X POST http://localhost:9999/structured \
  -d '{"prompt":"Return a JSON object with name, age, city for a fictional person"}'

# List available schemas
curl http://localhost:9999/structured/schemas

Built-in schemas: entities, summary, comparison, actions, classification, qa

Uses Gemini's native response_mime_type: application/json + responseSchema for guaranteed JSON output. Includes schema validation on the response.

Majority Voting (v1.3.7)

Same prompt → N parallel executions → pick the best answer. Higher accuracy on factual/analytical tasks.

# Judge strategy (LLM picks best — most reliable)
curl -X POST http://localhost:9999/vote \
  -d '{"prompt":"What are the key factors in SaaS pricing?","n":3,"strategy":"judge"}'

# Similarity strategy (consensus — zero extra cost)
curl -X POST http://localhost:9999/vote \
  -d '{"prompt":"What year was Python released?","n":3,"strategy":"similarity"}'

# Longest strategy (heuristic — zero extra cost)
curl -X POST http://localhost:9999/vote \
  -d '{"prompt":"Explain recursion","n":3,"strategy":"longest"}'

Strategies:

• judge — LLM scores all candidates on accuracy/completeness/clarity/actionability, picks winner (N+1 calls)
• similarity — Jaccard word-set similarity, picks consensus answer (N calls, zero extra cost)
• longest — Picks longest response as heuristic for thoroughness (N calls, zero extra cost)

When to use: Factual questions, critical decisions, or any task where accuracy > speed.

Self-Reflection (v1.3.5)

Optional critic pass after chain/skeleton output. Scores 5 dimensions, auto-refines if below threshold.

# Add reflect:true to any chain or skeleton request
curl -X POST http://localhost:9999/chain/auto \
  -d '{"task":"Analyze the AI chip market","data":"...","reflect":true}'

curl -X POST http://localhost:9999/skeleton \
  -d '{"task":"Write a market analysis","reflect":true}'

Proven: improved weak output from 5.0 → 7.6 avg score. Skeleton + reflect scored 9.4/10.

Skeleton-of-Thought (v1.3.6)

Generate outline → expand each section in parallel → merge into coherent document. Best for long-form content.

curl -X POST http://localhost:9999/skeleton \
  -d '{"task":"Write a comprehensive guide to SaaS pricing","maxSections":6,"reflect":true}'

Performance: 14,478 chars in 21s (675 chars/sec) — 5.1x more content than chain at 2.9x higher throughput.

When to use what:

• SoT → long-form content, reports, guides, docs (anything with natural sections)
• Chain → analysis, research, adversarial review (anything needing multiple perspectives)
• Parallel → independent tasks, batch processing
• Structured → entity extraction, classification, any task needing reliable JSON
• Voting → factual accuracy, critical decisions, consensus-building

API Endpoints

Cost Comparison

Cache hits are essentially free (~3-5ms, no API call).