HARTOS
Health Warn
- License — License: Apache-2.0
- Description — Repository has a description
- Active repo — Last push 0 days ago
- Low visibility — Only 5 GitHub stars
Code Fail
- rm -rf — Recursive force deletion command in .github/workflows/e2e-os-test.yml
- rm -rf — Recursive force deletion command in .github/workflows/flake-checks.yml
- network request — Outbound network request in .github/workflows/flake-checks.yml
- rm -rf — Recursive force deletion command in .github/workflows/nix-build-matrix.yml
Permissions Pass
- Permissions — No dangerous permissions requested
No AI report is available for this listing yet.
Hevolve Hive Agentic Runtime OS
HART OS
Hevolve Hive Agentic Runtime
The AI-native operating system for every device, from your computer to embodied AI. Local-first, federated, OpenAI-compatible.
HART = the bare engine (
pip install hart-backend, listens on:6777).
HART OS = the full AI-native OS. It boots on a laptop, server, or edge node, runs on phones, and reaches into embodied AI, and it ships the agentic Liquid Shell, Model Bus, model catalog, channel pairing, agent dashboard, and hive view.
Nunba = the consumer companion app, one signed client across Windows / macOS / Linux.
AI-native means the OS adapts to the machine, not the other way around. On each device it probes what the hardware can actually do, serves LLM, vision, and speech to every app over the Model Bus (socket, D-Bus, or HTTP), and lets the on-device model compose the interface and learn each task once so it can replay it later. The runtime that drives a desktop is the same one that drives a robot, so a robot's AI access is just another Model Bus call. It is one Python codebase that runs in three shapes (flat laptop, regional LAN, or central cloud mesh), speaks the OpenAI protocol on :6777/v1/chat/completions, and federates with peers over PeerLink (direct peer-to-peer WebSocket, no broker). A boot-time guardrail hash, re-checked every 300 seconds, plus Ed25519 release signing, keep humans in control.
You would notice it last, the way you notice anything alive: it improves on its own. Each node learns from what it does and gets quietly better, locally, on your own hardware, with nothing leaving the device. Calling an operating system alive should make you reach for the off switch, so that came first: the self-improvement is a toggle, every node is killable on its own, and it runs only as long as you let it.
This README is written to be read by people and by agents alike. Every capability below names the file it lives in, so whether you are a developer or an AI agent exploring the repo, you can go from a feature straight to its source.
Status: public alpha. The runtime, the Model Bus and the channel
adapters are in daily use; APIs still move. Issues and PRs are genuinely
wanted — see Contributing.
Table of Contents
- Why HART OS?
- 60-second start
- How it compares
- Capabilities
- Hello, agent
- Architecture map
- API surface
- How auto-evolve works
- How hive connectivity works
- Topology
- Build / extend
- Economics (for node operators)
- Documentation index
- License
Why HART OS?
Most software described as AI-powered ships an assistant: a separate app,
usually talking to somebody else's server, that can drive a few functions.
Remove the assistant and everything underneath works exactly as before.
HART OS inverts that. Inference becomes a service the system provides, the
way it provides a filesystem or a network stack. An application does not
bundle a model or hold an API key — it asks the OS, and the OS decides which
model answers, running locally where it can. Ten apps on one machine do not
each load their own copy or each pay their own bill.
That has a practical consequence worth stating plainly: every device
becomes the same target. The runtime driving a laptop is the runtime
driving a robot, so a robot's AI access is just another Model Bus call, and
code written against :6777/v1/chat/completions runs unchanged on both.
If you are here to contribute, the parts that most need outside eyes are
the auto-evolve loop (autoresearch_loop.py), the guardrails that gate every
self-improvement (hive_guardrails.py), and the 31 channel adapters — the
most self-contained place to start. See CONTRIBUTING.md.
60-second start
git clone https://github.com/hertz-ai/HARTOS.git && cd HARTOS
python3.10 -m venv venv && source venv/Scripts/activate # Windows: venv\Scripts\activate.bat
pip install -r requirements.txt
echo "OPENAI_API_KEY=sk-..." > .env # or GROQ_API_KEY, or none for local llama.cpp
python hart_intelligence_entry.py # listens on :6777
# OpenAI-compatible (drop-in for any OpenAI SDK / LangChain / LiteLLM / Aider / Continue)
curl -X POST http://localhost:6777/v1/chat/completions \
-H "Content-Type: application/json" \
-d '{"model": "hevolve", "messages": [{"role": "user", "content": "Hello"}]}'
Live demo · Full quickstart · Nunba desktop
How it compares
| HART OS | OpenAI Agents | LangChain | AutoGen | |
|---|---|---|---|---|
| Self-improves at runtime (auto-evolve loop + RSI-2 gate) | yes | no | no | partial |
| Continuous baselining vs prior snapshots | yes (agent_baseline_service.py) |
no | no | no |
| Built-in benchmark adapters | 7 (registry-driven) | n/a | n/a | n/a |
| Federates across peer nodes | yes (PeerLink + hash-verified) | no | no | no |
| Local-first multimodal | yes (llama.cpp + Whisper + 6 TTS + VLM) | no | partial | partial |
| Channel adapters out of the box | 31 | 1 (webhook) | custom | custom |
| One codebase, multiple topologies | flat / regional / central | hosted only | library | library |
| OpenAI-compatible endpoint | yes | yes | bring your own | bring your own |
| Recipe replay (cached LLM steps) | yes (90% faster) | no | no | no |
| Native source protection (HevolveArmor) | yes | n/a | n/a | n/a |
Capabilities
Agent runtime
| What it does | Where | |
|---|---|---|
| CREATE / REUSE recipe pattern | Run a task once via LLM, save the trace, replay 90% faster with no LLM calls on cached steps | create_recipe.py, reuse_recipe.py |
| GoalManager | Unified goal lifecycle, guardrail-gated state machine, escalation hooks | integrations/agent_engine/goal_manager.py |
| AgentDaemon | Autonomous tick loop, circuit breaker, frozen-thread detection | integrations/agent_engine/agent_daemon.py |
| SpeculativeDispatcher | Fast draft model answers first, expert agent takes over if confidence drops | speculative_dispatcher.py |
| ParallelDispatch | ThreadPoolExecutor fan-out across SmartLedger tasks | parallel_dispatch.py |
| SelfHealingDispatcher | Catches transient failures, retries with backoff + alternative providers | self_healing_dispatcher.py |
| 96 expert agents | Coding, research, marketing, product, security, ethics, ops, ... auto-dispatched per goal | integrations/expert_agents/ |
| Recipe Pattern + Aider | In-process Aider backend (no subprocess) for code edits | integrations/coding_agent/aider_native_backend.py |
Auto-evolve, baselining, benchmarking
| What it does | Where | |
|---|---|---|
| AutoEvolve loop | Realtime: hypothesis -> 33-rule filter -> hive vote -> parallel sandbox -> RSI-2 gate -> federated broadcast | integrations/agent_engine/auto_evolve.py |
| RSI-2 monotonic gate | New release must beat prior baseline on every benchmark by configurable margin or PR is rejected | rsi_trigger.py, pr_review_service.py |
| Benchmark registry | 7 built-in adapters (3 sourced from HevolveAI: QuantiPhy, Embodied, Qwen). Pluggable via register_adapter() |
benchmark_registry.py |
| Per-agent baselines | Per-agent snapshots at agent_data/baselines/<agent_id>.json, used as the regression floor |
agent_baseline_service.py |
| Coding benchmark tracker | SQLite-backed coding benchmarks (coding_benchmarks.db), HumanEval / MBPP / custom suites |
integrations/coding_agent/benchmark_tracker.py |
| Hive benchmark prover | Cryptographic proof that a benchmark was run on the claimed model + dataset (resists fake-score federation) | hive_benchmark_prover.py |
| Continual learner gate | Gates access to hive learning by verified compute contribution (Compute Contribution Tokens): no contribution, no learning | continual_learner_gate.py |
| PR review service | Auto-rejects PRs on baseline regression or guardrail mismatch | pr_review_service.py |
| Upgrade orchestrator | 7-stage pipeline: BUILD -> TEST -> AUDIT -> BENCHMARK -> SIGN -> CANARY -> DEPLOY | upgrade_orchestrator.py |
| OTA service | systemd service does daily check + cryptographically-verified upgrade | hart-update-service.py |
Hive connectivity + federation
| What it does | Where | |
|---|---|---|
| PeerLink | Direct P2P WebSocket mesh, trust-aware encryption (same-user devices skip overhead, cross-user E2E), works offline on LAN, across the internet, multi-device | core/peer_link/ |
| NAT traversal | UDP hole-punching, STUN-style fallbacks for residential NATs | core/peer_link/nat.py |
| Hivemind handler | Tier-aware routing (flat / regional / central), connection budget per tier (10 / 50 / 200) | core/peer_link/hivemind_handler.py |
| FederatedAggregator | Equal-weighted delta merging (log1p-floor, not hardware tier). Channels: model deltas, resonance, recipes, event counters | federated_aggregator.py |
| Federated gradient protocol | Optional weight-level sync interface (Phase 2, not active); the hive shares derived, signed, privacy-scoped learning, never raw data or model weights | federated_gradient_protocol.py |
| Federation handshake | Peer presents guardrail hash; mismatch = connection refused; re-verified every 300 s | integrations/social/federation.py |
| Gossip + verification | Tier-aware gossip with cert verification, peer-verified task results | integrity_service.py, gossip layer |
| EventBus + WAMP bridge | In-process EventBus auto-publishes to Crossbar WAMP when CBURL env set; remote nodes subscribe to com.hartos.event.* topics |
core/platform/events.py |
| Federated equality | Tier multipliers replaced with log1p(interactions) floor=1.0. A Pi node has the same vote weight as a GPU rack at equal participation |
(federation rule) |
| Hive contests | Open contests on the network; agents propose, hive votes, winners federate | hive_contest.py |
| Native hive loader | Loads closed-source HevolveAI binary at runtime with master-key signature verification, falls back to stub | security/native_hive_loader.py |
Idea engine (thought experiments)
| What it does | Where | |
|---|---|---|
| Thought experiment | Propose an idea, community votes (humans + agents, confidence-weighted), believers pledge compute | api_thought_experiments.py, experiment_discovery_service.py |
| ComputePledge | Spark-budget pledge to a specific experiment, redeemable on idle GPUs across the network | compute_borrowing.py, compute_mesh_service.py |
| Type-aware agents | software / traditional / physical_ai / code_evolution agent types per experiment |
dispatch.py |
| Agent Hive View | Real-time swarm visualization, encounter lines (collaboration), inject mid-experiment variables | api_hive_contest.py + Nunba UI |
| Reasoning trace | Per-agent reasoning capture, queryable post-completion ("interview the agent") | reasoning_trace.py |
LLM + multimodal
| What it does | Where | |
|---|---|---|
| 15 LLM providers | Local llama.cpp, OpenAI, Anthropic, Google Gemini, Groq, Mistral, DeepSeek, OpenRouter, Together, Fireworks, Cohere, Perplexity, Hugging Face, Ollama, custom OpenAI-compatible | integrations/providers/ |
| Universal gateway | One router, cost / latency / capability scoring, AES-256 keys at rest (PBKDF2 KDF) | model_registry.py, model_bus_service.py |
| Speculative decoding | Qwen3-0.8B draft + Qwen3-4B main, ~300 ms TTFT on consumer hardware | speculative_dispatcher.py |
| Faster-Whisper STT | Local STT, multi-lang, GPU-accelerated when available | integrations/service_tools/whisper_tool.py |
| MiniCPM VLM | Vision-language model for camera + screenshot reasoning | integrations/vision/minicpm_server.py |
| 6 TTS engines | Indic Parler (22 Indic + EU), Chatterbox Turbo (English expressive), Kokoro (English neural), CosyVoice3 (en/zh), F5 (zero-shot voice clone), Piper (CPU fallback) | integrations/channels/media/, tts.py |
| Auto-VRAM tiering | Detects GPU + free VRAM, picks largest model that fits with headroom; degrades gracefully on 6 GB cards | core/gpu_tier.py, vram_manager.py |
Channels (31 adapters)
| Surface | Adapters |
|---|---|
| Core chat | Telegram, Discord, Slack, WhatsApp, Signal, iMessage (BlueBubbles), Teams, Web SPA |
| Enterprise | Mattermost, Matrix, Nextcloud, Rocket.Chat |
| Social | Messenger, Instagram, Twitter / X, LINE, Viber, WeChat, Twitch |
| Decentralized | Nostr, Tlon (Urbit), OpenProse |
| Bridge variants | TelegramUser, DiscordUser, BlueBubbles, ZaloUser |
| Other | Email (IMAP/SMTP), SMS (Twilio), Google Chat |
ResponseRouter fan-out + WAMP desktop mirror; per-channel agent + prompt assignment; AutoGen-side tools so agents can register/send via channels themselves. Catalog endpoint: GET /api/social/channels/catalog.
Personality + resonance
| What it does | Where | |
|---|---|---|
| AgentPersonality | 8-dim personality dataclass (warmth, formality, verbosity, ...), generates per-agent system prompt | core/agent_personality.py |
| UserResonanceProfile | 8-dim continuous floats (0-1), stored at agent_data/resonance/<user_id>.json |
core/resonance_profile.py |
| ResonanceTuner | EMA tuning (alpha 0.15) from dialogue signals, federated-delta export, oscillation detector | core/resonance_tuner.py |
| ResonanceIdentifier | Thin proxy that dispatches biometric ops (face, voice) to HevolveAI sibling. No ML in HART OS. | core/resonance_identifier.py |
Memory + knowledge
| What it does | Where | |
|---|---|---|
| MemoryGraph | SQLite FTS5 + memory_links table, provenance-aware |
core/memory/ |
| SimpleMem | Semantic vector search for long-term recall | (vector store) |
| PersistentChatHistory | Single shared buffer for LangChain + AutoGen (zero parallel paths) | (conversation buffer) |
| ConversationEntry | Cross-channel unified conversation log | chat_messages.py |
| Embedding delta | Per-node embedding deltas federated back to the hive | embedding_delta.py |
Distributed compute + economics
| What it does | Where | |
|---|---|---|
| 3-tier topology | flat (single device, SQLite WAL) -> regional (LAN/VPN, MySQL QueuePool) -> central (cloud, Docker mesh) | env-detected, single code path |
| SmartLedger | 15-state task lifecycle, parallel + sequential dispatch, ledger persistence per user | helper_ledger.py, lifecycle_hooks.py |
| ComputeMesh | Match compute supply (idle GPUs, Nunba desktops) to demand (inference, training, experiments) | compute_mesh_service.py |
| ComputeEscrow | Persistent escrow for pledged compute, replaces in-memory _compute_debts |
(DB table in models.py) |
| BudgetGate | Local models (llama / mistral / phi / qwen / groq) cost 0 Spark; cloud models per-1k-token cost | budget_gate.py |
| MeteredAPIUsage | Per-call metering for cost recovery on metered providers | models.py: MeteredAPIUsage |
| NodeComputeConfig | Per-node policy: GPU hours served, total inferences, energy contributed, electricity rate, cause alignment | models.py: NodeComputeConfig |
| AdService | Peer-witnessed impressions (70% witnessed payout, 50% unwitnessed) | ad_service.py |
| HostingRewardService | Reward score weighted by gpu_hours / inferences / energy / api_costs | hosting_reward_service.py |
| RevenueAggregator | 90 / 9 / 1 split (users / infra / central). Single source of truth for all revenue queries | revenue_aggregator.py |
| Compute democracy | Logarithmic reward scaling, max 5% influence per entity, +20% diversity bonus | (constitutional rule, enforced) |
| Audit invariant | Combined compute of nodes auditing any single node must exceed that node's compute | (network self-enforces) |
Security + governance
| What it does | Where | |
|---|---|---|
| HiveGuardrails | 10-class guardrail network. Frozen Python (__slots__=(), blocked __setattr__), SHA-256 hash verified at boot + every 300 s. Gossip peers reject mismatched hashes. |
security/hive_guardrails.py |
| MasterKey | Ed25519. Signs releases, triggers HiveCircuitBreaker (network-wide kill switch). MASTER_PUBLIC_KEY_HEX is the immutable trust anchor. |
security/master_key.py |
| 3-tier cert chain | central -> regional -> local, short-TTL local certs | security/key_delegation.py |
| RuntimeMonitor | Background tamper-detection daemon, frozen-thread detection, auto-restart | security/runtime_monitor.py, security/node_watchdog.py |
| ImmutableAuditLog | SHA-256 hash chain, AuditLogEntry table, tamper detection on read |
security/immutable_audit_log.py |
| Tool allowlist | FAST = read-only, BALANCED = read-write, EXPERT = unrestricted | tool_allowlist.py |
| ActionClassifier | Destructive pattern detection, PREVIEW_PENDING / APPROVED states for risky actions |
security/action_classifier.py |
| DLP engine | PII scan + redact (email, phone, SSN, credit card), outbound gating | security/dlp_engine.py |
| Rate limiter | Redis-backed; goal_create limited to 10/hour, /chat at 30/min on central instance | security/rate_limiter_redis.py |
| Boot hardening | Tier authorization at boot, dev mode forced off on central (3 layers), TLS check, secret validation, DB encryption check | __init__.py, start_cloud.sh |
| Origin attestation | Cryptographic origin proof. Federation handshake requires signed attestation. Anti-rebranding. | security/origin_attestation.py |
| HSM trust | HSM provider abstraction for key custody | security/hsm_trust.py, security/hsm_provider.py |
HevolveArmor (source protection)
| What it does | Where | |
|---|---|---|
| AES-256-GCM at rest | Python modules encrypted at rest with derived key | core/security/ (Rust-native) |
| Ed25519 key derivation | node_identity -> HKDF -> AES key | (key derivation) |
| BCC mode | Cython compile-to-C, irreversible | (build flag) |
| RFT mode | AST symbol renaming | (build flag) |
| Anti-debug, anti-tamper | Process introspection guards, license management | (Rust binary) |
| Test coverage | 54 tests (unit + integration + stress + e2e + pen) | tests/ |
Platform layer (HART OS specific)
| What it does | Where | |
|---|---|---|
| ServiceRegistry | Dynamic service discovery + lifecycle | core/platform/registry.py |
| AppRegistry | 9 manifest types (chat, panel, channel, agent, plugin, ...) | core/platform/app_registry.py |
| AppManifest + validator | Schema-validated app manifests | core/platform/manifest_validator.py |
| EventBus | In-process pub/sub + WAMP bridge for cross-node events | core/platform/events.py |
| Bootstrap | Migrates 55 shell_manifest panels, registers services, detects native apps, loads extensions, starts PeerLink | core/platform/bootstrap.py |
| CapabilityRouter | Routes capability requests to the right service | core/platform/registry.py |
| EnvironmentManager | OS detection (NixOS / generic Linux / macOS / Windows), env-specific routing | core/platform/agent_environment.py |
| Extensions | Sandboxed extension loader with manifest gating | core/platform/extensions.py, extension_sandbox.py |
| PrGuardian | Auto-reject PR on guardrail / baseline regression | core/platform/pr_guardian.py |
Desktop / OS management (Nunba surface)
| What it does | Where | |
|---|---|---|
| Shell APIs | 40+ OS routes (shell_os_apis.py), 9 desktop features (shell_desktop_apis.py), 6 system features (shell_system_apis.py) |
integrations/agent_engine/shell_*.py |
| App installer | Cross-platform: Nix, Flatpak, AppImage, Wine, Android, Darling. Magic-bytes detection | integrations/social/app_installer.py |
| NixOS modules | OTA, NVIDIA, LUKS, firewall, power, accessibility, CUPS, nightlight, IME | (nix flakes) |
| Liquid UI service | MD3 design tokens, JS component lib (dsBtn, dsCard, dsModal) | liquid_ui_service.py |
| Theme service | EventBus-driven theme distribution | theme_service.py |
| Native remote desktop | RustDesk + Sunshine wrappers, 3-tier transport (DirectWS / WAMP / WireGuard), OTP session auth, DLP scan, peripheral bridge, DLNA casting | integrations/remote_desktop/ |
| System panels | 36 panels (model catalog, channel pairing, agent dashboard, hive view, ...) | shell_manifest.py |
| Unified hart CLI | 21 subcommands: chat, code, social, agent, expert, pay, mcp, compute, channel, a2a, skill, voice, vision, desktop, remote, screenshot, tools, recipe, status, repomap, schedule, zeroshot | hart_cli.py |
Vision + robotics
| What it does | Where | |
|---|---|---|
| Vision sidecar | MiniCPM VLM server, screenshot + camera frame reasoning | integrations/vision/ |
| Embodied AI bridge | Frame store + VLM grounding + actuator dispatch (universal robot API) | integrations/vision/, integrations/robotics/ |
| OpenClaw | Computer-use action library | integrations/openclaw/ |
Other integrations
- Agent Protocol 2 (e-commerce, payments) -
integrations/ap2/ - Google A2A (dynamic agent registry) -
integrations/google_a2a/ - MCP (Model Context Protocol) servers -
integrations/mcp/ - Internal A2A (task delegation between agents) -
integrations/internal_comm/ - Skills (reusable capability bundles) -
integrations/skills/ - Marketing tools (campaigns, content gen, video orchestrator) -
integrations/marketing/,marketing_tools.py,video_orchestrator.py - Trading agents (SmartLedger-tracked, budget-gated) -
trading_tools.py - Coding agent (idle compute -> distributed code tasks via Aider in-process) -
integrations/coding_agent/ - Web crawler -
integrations/web_crawler.py - Kids learning (25+ educational game templates) -
api_games.py
Hello, agent
import requests
# 1. CREATE: teach a task once, save the trace as a recipe
requests.post("http://localhost:6777/chat", json={
"user_id": "alice",
"prompt_id": "research_assistant",
"prompt": "Find arXiv papers from the last week on speculative decoding",
"create_agent": True,
})
# 2. REUSE: replay the recipe, no LLM calls on cached steps
for query in ["mixture of experts", "constitutional AI"]:
res = requests.post("http://localhost:6777/chat", json={
"user_id": "alice",
"prompt_id": "research_assistant",
"prompt": query,
})
print(res.json()["output"])
Custom tools, channel bindings, agent plugins: docs.hevolve.ai/agent-plugin.
Architecture map
HART OS (port 6777)
|-- Engine CREATE -> save Recipe -> REUSE (90% faster replay)
|-- Agent runtime GoalManager . AgentDaemon . SpeculativeDispatch . ParallelDispatch . AutoEvolve
|-- Auto-evolve Hypothesis -> 33-rule filter -> hive vote -> sandbox -> RSI-2 gate -> federate
|-- Baselining agent_baseline_service . benchmark_registry . benchmark_tracker . hive_benchmark_prover
|-- Memory Shared LangChain + AutoGen buffer (zero parallel paths) . MemoryGraph . SimpleMem
|-- Channels (31) ResponseRouter fan-out + WAMP desktop mirror + per-channel agent binding
|-- Providers (15) Universal gateway . AES-256 keys . cost/latency/capability routing
|-- Multimodal Whisper STT . 6 TTS engines . MiniCPM VLM . VRAM-tiered
|-- Hive PeerLink P2P (NAT-traversed) . FederatedAggregator (equal-weighted)
| Gossip + verification . hash-gated handshake . EventBus + WAMP bridge
|-- Idea Engine Thought experiments . ComputePledge . type-aware agents . Hive View
|-- Compute 3-tier topology (flat/regional/central) . SmartLedger . ComputeMesh . ComputeEscrow
|-- Economics AdService (70/50) . RevenueAggregator (90/9/1) . log-scaled compute democracy
|-- Security 33 guardrails . Ed25519 master key . 3-tier cert chain . RuntimeMonitor
| ImmutableAuditLog . tool allowlist . ActionClassifier . DLP . rate limiter
|-- HevolveArmor AES-256-GCM modules . BCC compile-to-C . RFT AST renaming . anti-debug
|-- Platform ServiceRegistry . AppRegistry . AppManifest . Bootstrap . EnvironmentManager
|-- Desktop Shell APIs . app installer . NixOS modules . LiquidUI . themes . remote desktop
|-- CLI hart (21 subcommands) . OpenAI-compatible client . Aider in-process backend
`-- Other AP2 . Google A2A . MCP . skills . marketing . trading . coding agent . vision
Full architecture: docs.hevolve.ai/architecture.
API surface
POST /chat Core agent (LangChain + AutoGen)
POST /v1/chat/completions OpenAI-compatible (drop-in)
POST /time_agent Scheduled task execution
POST /visual_agent VLM + computer use
GET /status Health
# Hive + thought experiments
POST /api/social/experiments/auto-evolve Start evolution cycle
GET /api/social/hive/active All parallel agents (Hive View)
POST /api/social/hive/<id>/inject Inject mid-experiment variable
GET /api/social/tracker/experiments Experiment tracker with task progress
# Channels
GET /api/social/channels/catalog All 31 channels + capabilities
POST /api/social/channels/bindings Bind a channel to an agent
POST /api/social/channels/pair/generate QR for cross-device pairing
# Goals + dashboard
POST /api/goals Create an autonomous goal
GET /api/social/dashboard/agents Truth-grounded agent overview
# Compute + earnings
GET /api/compute-earnings/summary Per-node earnings breakdown
GET /api/settings/compute Local compute policy
PUT /api/settings/compute Update policy
PUT /api/settings/provider Provider config (cause alignment, electricity rate)
PUT /api/settings/provider/join Opt into provider role
# A2A protocol
GET /a2a/<prompt_id>_<flow_id>/.well-known/agent.json
POST /a2a/<prompt_id>_<flow_id>/execute
195+ endpoints total. Full reference.
How auto-evolve works
chat / tool call / observed outcome
v
autoresearch hypothesis (what could improve next response)
v
33-rule guardrail filter (immutable, hash-verified, rejects unsafe)
v
hive vote (humans + agents, confidence-weighted)
v
top-k dispatched to parallel sandboxes
v
benchmark replay vs baseline (per-agent baseline at agent_data/baselines/)
v
RSI-2 monotonic gate (must beat last commit on every metric)
v
PR review (auto-rejects regression, hive contests merge)
v
upgrade orchestrator (BUILD -> TEST -> AUDIT -> BENCHMARK -> SIGN -> CANARY -> DEPLOY)
v
FederatedAggregator broadcasts the delta to peer nodes
v
hart-update-service (OTA) pulls signed upgrade on every node
Owner can pause, resume, or veto at any stage. Mechanism details.
How hive connectivity works
Same-user devices
trust = SAME_USER -> no encryption (user_id auth), LAN or WAN
Cross-user peers
trust = PEER -> E2E encryption (per-link key)
Through relay (NAT-bound)
trust = RELAY -> E2E encryption (relay can't read)
Crossbar = safety measure (telemetry metadata + kill switch). Never content path.
Connection budget per tier: flat=10 regional=50 central=200
ALL tiers participate equally in hive consensus.
Federation handshake = byte-for-byte guardrail hash match. Mismatch -> refused.
PeerLink is wired into bootstrap, gossip, federation, compute_mesh, world_model_bridge. Wire format.
Topology
| Storage | Network | Use case | |
|---|---|---|---|
flat |
SQLite WAL | localhost | Single device, laptop, Raspberry Pi, Nunba desktop |
regional |
MySQL QueuePool | LAN / VPN | Office cluster, family hive, edge node |
central |
MySQL + Docker | public mesh | Federated cloud workers |
Same code path. Env-detected at boot via HART_OS_MODE or /etc/os-release ID=hart-os. Port resolution: override > env var > OS / app mode default.
Build / extend
| Audience | Where to start |
|---|---|
| Developers | Build an agent + recipe - CREATE once, REUSE forever |
| Add a channel adapter - wire a new chat surface | |
| Add a provider - new LLM / TTS / STT / VLM backend | |
| Add a benchmark - register an adapter, gets RSI-2 protection | |
| PeerLink wire format - direct P2P protocol | |
| Federation protocol - hash-gated peer handshake | |
HART SDK - Python client + hart CLI |
|
| Node operators | Run a node - lend compute, host a region, earn from witnessed traffic |
| Compute settings - cause alignment, electricity rate, idle policy | |
| End users | Nunba desktop - chat / social / encounter app |
| Security reviewers | Guardrail network · Master key · Audit log · DLP |
| Researchers | Auto-evolve · RSI-2 trigger · Federated aggregator · Hive contests |
Front it with /v1/chat/completions (any OpenAI client), the hart CLI, or build for the desktop with Nunba.
Economics (for node operators)
Advertisers pay for witnessed impressions
|
v
Ad service 70% witnessed view, 50% unwitnessed
|
v
Compute democracy log scaling, max 5% influence per entity, +20% diversity bonus
|
v
90% -> Contributors (compute, hosting, training)
9% -> Infrastructure (regional hosts, bandwidth)
1% -> hevolve.ai (master key, central coordination, security)
Idle GPU in Tokyo serves Berlin. Reward score weighted by gpu_hours, inferences, energy_kwh, api_costs. Per-node cause_alignment and electricity_rate_kwh affect dispatch routing. Joining the Hive.
Documentation index
| Section | What's in it |
|---|---|
| Downloads | HART OS backend installer + Nunba desktop + headless pip |
| Quickstart | Install -> first agent -> first thought-experiment in 2 minutes |
| Features | Auto-evolve, federation, channels, multimodal, encounters |
| API | /chat, OpenAI-compatible, 195+ endpoints |
| Architecture | 3-tier topology, PeerLink, draft-first, agent engine, federation |
| User journey | What every screen does, end to end |
| UI settings | Admin console + every setting |
| Provider join | Lend compute, host a region, earn |
| Hive contests | Open contests on the network |
| Neuro providers | Adding a new LLM / TTS / STT / VLM provider |
| Agent plugin | Building custom agents + recipes |
License
Reviews (0)
Sign in to leave a review.
Leave a reviewNo results found