LatticeAG PolyBrain 🧠

A multi-agent, multi-model orchestration layer for Hermes Agent.
Parallel research. Clean synthesis. Verified claims.

Quick Start · Why PolyBrain · How It Works · Features · Configuration

---

PolyBrain turns a single objective into a coordinated multi-agent workflow: it decomposes the task, routes work to specialized roles, executes subtasks in parallel, synthesizes the outputs, and verifies every claim against cited sources.

Inspired by the orchestration pattern behind Perplexity Computer - built as a local, config-driven, reproducible skill with a polished, high-trust workflow.

Why PolyBrain

• Designed for high-signal work - built to keep outputs focused, sourced, and useful.
• Parallel by default - research and build tasks run concurrently to reduce wait time.
• Source-first - claims must be backed by citations, not vibes.
• Verification built in - final output is checked claim-by-claim before it reaches you.
• Multi-model by design - route different roles to different models/providers in config.yaml.

How PolyBrain is different

• Multi-model, not just multi-agent - different roles can use different LLMs, not just different agents on the same model. Unlike Hermes's native delegate_task and Kanban - which spin up subagents but bind them all to the same model config - PolyBrain routes each role independently. Run DeepSeek for research, Claude for synthesis, GPT-4o for verification. Same pipeline, genuinely different brains.
• Verification layer - claims are checked against sources before the final answer is returned.
• Citation enforcement - uncited claims are dropped, not silently included.

Quick Start

# 1. Clone and install
git clone --depth=1 https://github.com/mosesman831/PolyBrain.git /tmp/polybrain
rm -rf /tmp/polybrain/.git
cp -r /tmp/polybrain ~/.hermes/skills/research/polybrain
rm -rf /tmp/polybrain

# 2. Edit config.yaml with your model aliases
#    (orchestrator, researcher, builder, synthesizer, verifier, fallback)
hermes config edit  # then edit ~/.hermes/skills/research/polybrain/config.yaml

# 3. Validate config
python ~/.hermes/skills/research/polybrain/scripts/validate_config.py

# 4. Use it - just tell Hermes what you want in a chat:
"Use PolyBrain to research Apple's latest earnings and competitors"
# Hermes will load the skill and run the orchestration script for you.

For advanced/manual use, you can also run the script directly:

echo "Summarize Apple's latest quarterly earnings with sources" | \
  python ~/.hermes/skills/research/polybrain/scripts/orchestrate.py

How It Works

flowchart TD
  A[Objective] --> B[Orchestrator]
  B --> C1[Researcher 1]
  B --> C2[Researcher 2]
  B --> C3[Builder]
  C1 --> D[Synthesizer]
  C2 --> D
  C3 --> D
  D --> E[Verifier]
  E --> F[Final Answer]

Loading

Features

• Premium orchestration flow - clean task splitting, parallel execution, and final synthesis.
• Citation enforcement - researchers must include URLs; uncited claims are dropped.
• Source verification - the verifier checks each claim against its source and returns PASS/FAIL.
• Robust JSON parsing - orchestrator output is recovered even when models add extra prose.
• Retry logic - transient failures are retried before the run is marked failed.
• Role-based routing - assign different models/providers per role in config.yaml.
• Artifact logging - every run is saved in a timestamped folder for traceability.

Roles

Role	Purpose	Toolsets
Orchestrator	Decompose objective into a JSON task plan	text-only
Researcher	Web search + citations	web, browser
Builder	Code, terminal, and file operations	terminal, file
Synthesizer	Merge outputs into the final deliverable	optional
Verifier	Verify claims against cited sources	web

Example

Tell Hermes:

Use PolyBrain to research Apple's latest quarterly earnings and competitors

Or run the script directly:

echo "Summarize Apple's latest quarterly earnings with sources" | \
  python ~/.hermes/skills/research/polybrain/scripts/orchestrate.py

What you get

• A decomposed task plan
• Parallel research with citations
• A synthesized final brief
• Claim-by-claim verification

Configuration

Edit config.yaml:

models:
  orchestrator: "your-model"
  researcher: "your-model"
  builder: "your-model"
  synthesizer: "your-model"
  verifier: "your-model"
  fallback: "your-model"

providers:
  orchestrator: "your-provider"  # optional, set per-role if needed
  researcher: "your-provider"

settings:
  max_parallel: 3
  timeout_sec: 300
  orchestrator_timeout_sec: 120
  artifacts_dir: ".hermes/plans/polybrain"

See config.yaml for all options.

Learn More

• Architecture
• Prompt Templates
• JSON Schema
• JSON Parsing Strategy
• Example Run

File Tree

polybrain/
├── SKILL.md                          # Skill definition (Hermes)
├── README.md                         # This file
├── config.yaml                       # User-editable model configuration
├── references/
│   ├── architecture.md               # Design docs + data flow
│   ├── prompts.md                    # Versioned prompt templates
│   ├── schema.json                   # JSON schema for orchestrator output
│   ├── json-output-parsing.md        # JSON robustness strategy
│   └── example-run.md                # Example objective → final output
└── scripts/
    ├── orchestrate.py                # Production runner (parallel + sequential)
    ├── orchestrate_debug.py          # Debug runner (all sequential, verbose)
    └── validate_config.py            # Config validator

Known Issues

• Model-specific hangs - Some models (e.g. gpt-5-mini via certain providers) can hang in hermes chat subagent calls. If a model hangs for 300s+, try a different model or provider. Test with hermes chat -q "ping" -m your-model first.
• Verifier may truncate numbers - Some models strip leading digits from dollar amounts in verification reports. The PASS/FAIL verdicts remain structurally valid.

License

GNU General Public License version 3