▲ {{ currentViewBreadcrumb }}
▲ COGNITIVE & TOKEN FLOW TOPOLOGY LIVE ROUTING TO OMNIROUTER & SANDBOX
TELEGRAM:
ALERTMANAGER:
VS CODE PROVIDER
/v1/chat
● READY
TELEGRAM BOT
@hal9k_bot
{{ channels.telegram ? '● LISTENING' : '○ MUTED' }}
HAL9k COGNITIVE CORE
{{ activeProvider ? activeProvider.name : 'OMNIROUTER GATEWAY' }}
AES-256 VAULT // BLAKE3
DOCKER SANDBOX
1 CPU / 4GB RAM
BOLLARD RUNNER
REMOTE INFRA (SSH)
RUSSH SCREEN
MAX 5 SESSIONS
▲ SYSTEM TELEMETRY CYCLE: 30s
Cognitive Confidence: {{ (systemState.governor_confidence * 100).toFixed(0) }}% NOMINAL
Active Scheduled Heartbeats: {{ systemState.active_heartbeats }} / {{ heartbeats.length }}
Installed Skills / Runbooks: {{ systemState.total_skills }} SKILLS
Configured AI Providers: {{ providers.length }} PROVIDERS
AE-35 Comm Unit Status: {{ systemState.ae35_status }}
▲ GOVERNOR CIRCUIT CHIP BLAKE3 DISJUNCTOR
ACTION HASH GATE ACTIVE (BLAKE3)
LOOP DISJUNCTOR ARMED (< 0.40)
BLAST RADIUS GUARD SAFE-MODE ENFORCED
TIMEOUT SUICIDE SIGKILL -- 5s
▲ MISSION TERMINAL OUTPUT
[{{ log.time }}] {{ log.text }}
DISCOVERY ONE
HAL 9000
GOVERNOR CONFIDENCE DECAY
{{ (systemState.governor_confidence * 100).toFixed(0) }}%
LOGICAL POWER SWITCH
PANIC DISCONNECT (/state/kill)
▲ ACTIVE CONFIGURATION
OMNIROUTER ENDPOINT:
{{ settings['omnirouter.url'] || 'https://omni.highdevs.com.br' }}
SUMMARIZER COMPRESSOR MODEL:
{{ settings['summarizer.model'] || 'gpt-4o-mini' }}
MAX TOKENS PER TASK:
{{ settings['agent.max_tokens_per_task'] || '4096' }} TOKENS
PROMETHEUS TELEMETRY: /metrics

SRE MISSION TERMINAL & RUNNER

Dispatch natural language infrastructure & engineering tasks to HAL9k. Tools, Docker sandbox, and SSH are orchestrated autonomously.

MODEL:
MAX REASONING:
QUICK PRESETS:
HAL9k COGNITIVE CORE AT WORK: Evaluating task, calling tools, and enforcing Governor disjunctor...
ITERATION ACTIVE
▲ FINAL AGENT RESOLUTION STATUS: {{ currentMissionResult.status }}
{{ currentMissionResult.total_duration_ms }} ms
{{ currentMissionResult.final_answer }}
▲ STEP-BY-STEP TOOL EXECUTION TRACE ({{ currentMissionResult.steps.length }} STEPS)
No steps match the selected filter.
STEP #{{ step.step }} {{ step.action }}
GOVERNOR CONFIDENCE: {{ (step.confidence * 100).toFixed(0) }}% DURATION: {{ step.duration_ms }}ms
ARGUMENTS: 
{{ JSON.stringify(step.input, null, 2) }}
OBSERVATION / TOOL OUTPUT: 
{{ step.output }}

AGENT TASKS & LIFECYCLE MANAGEMENT

Track multi-agent missions, inspect exact system prompts used in each execution, re-run with stored prompts, or cancel active tasks.

STATUS TASK CODE ORIGIN SUB-AGENT USER PROMPT DURATION ACTIONS
{{ t.status }} {{ t.task_code }} {{ t.origin }} {{ t.subagent_role }} {{ t.prompt }} {{ t.duration_ms }} ms
No agent tasks match the selected criteria. Missions dispatched from VS Code, Web Terminal, or Telegram appear here.

WORKSPACE FILE EXPLORER

Inspect, create, and edit files, generated code, and incident logs in the HAL9k workspace filesystem.

TYPE FILENAME PATH SIZE ACTION
{{ f.is_dir ? 'DIR' : 'FILE' }} {{ getFileIcon(f.name, f.is_dir) }} {{ f.name }} {{ f.path }} {{ f.is_dir ? '-' : formatFileSize(f.size_bytes) }}
No files found in workspace directory. Click "NEW FILE" to create artifacts or run a mission.

TELEGRAM SENTRY & ACCESS CONTROL

Inspect Telegram interactions, approve or revoke chat access, and configure dedicated personas/models per user.

ACCESS CHAT ID USERNAME FIRST NAME ASSIGNED MODEL ROLE LAST SEEN ACTION
{{ u.chat_id }} {{ u.username ? '@' + u.username : 'Anonymous' }} {{ u.first_name || '-' }} {{ u.role }} {{ u.last_seen_at.split(' ')[0] }}
No Telegram users recorded yet. Send a message to your configured bot on Telegram to log the user session here.

VISUAL DAG PIPELINES (DIFY / N8N ENGINE)

Design autonomous multi-agent pipelines with conditional branching, LLMs, Docker sandboxes, MCP tools and channel outputs.

🖱️ Arraste o nó para mover | 🔌 Puxe do ● OUT para ligar no IN | ⚙️ Clique no card para editar
Clique no fio para remover a conexão
{{ getNodeIcon(node.type) }} {{ node.id }}
{{ node.type }}
{{ getNodeLabel(node) }} EDITAR ✎
{{ getNodeDetails(node) }}
IN
START
TRUE
FALSE
OUT
END
EXECUTION TRACE: {{ lastPipelineRun.status }} ({{ lastPipelineRun.duration_ms }}ms)
Run ID: {{ lastPipelineRun.run_id }}
{{ step.node_id }} [{{ step.node_type }}] {{ step.status }}
{{ JSON.stringify(step.output).slice(0, 160) }}
{{ step.duration_ms }}ms

SHORTCUTS, SNIPPETS & MACRO DIRECTORY

Directory of prompt macros, pipeline variables, MCP triggers, and HITL directives usable across HAL9k.

🔍
{{ item.category }}

{{ item.name }}

{{ item.syntax }}

{{ item.desc }}

EXAMPLE PROMPT:
{{ item.example }}
No shortcuts found matching "{{ shortcutSearch }}"

MODEL CONTEXT PROTOCOL (MCP) REGISTRY

Install and manage dynamic MCP tools (Postgres, GitHub, Filesystem, Custom CLI tools) usable by HAL9k during missions.

MCP PROTOCOL

{{ s.name }}

EXECUTABLE COMMAND: {{ s.command }} {{ s.args.join(' ') }}
INSTALLED BY: {{ s.installed_by }}
{{ s.created_at.split(' ')[0] }}
🔌
No MCP Servers Registered

Install an MCP server or ask the agent during a mission (e.g. "Install Postgres MCP") to register tools.

LUA PLUGINS & FUNCTION EXTENSIONS

Extensões dinâmicas em Lua 5.4 isoladas em sandbox com memory cap (32MB) e ponte nativa hal9k. Instale pacotes .zip.

🌙 LUA 5.4 // v{{ p.version }} ID: {{ p.id }}

{{ p.name }}

{{ p.description || 'Nenhuma descrição fornecida.' }}

Entrypoint: {{ p.entrypoint || 'plugin.lua' }}
Tools Exportadas: {{ p.tools ? p.tools.length : 0 }}
🛡️ 32MB CAP ⏱️ 2M CPU HOOK ⚡ 15s TIMEOUT
Ferramentas Disponíveis: Clique para testar
⚡ {{ t.name }}
ATIVO NO ORCHESTRATOR
📦
NENHUM PLUGIN LUA INSTALADO

Faça upload de um pacote .zip contendo manifest.json e plugin.lua para adicionar novas ferramentas em runtime.

AI MODEL PROVIDERS

Configure connection endpoints to OmniRouter Gateway, OpenAI, Anthropic, or custom inference nodes.

{{ prov.provider_type }} ● ACTIVE

{{ prov.name }}

BASE URL: {{ prov.base_url }}
DEFAULT INFERENCE MODEL: {{ prov.default_model }}
{{ prov.created_at.split(' ')[0] }}
📡
No AI Providers Configured

Register OmniRouter Gateway or OpenAI credentials to activate the agent's inference engine.

DUAL HEARTBEATS ENGINE

Configure Cognitive Loop interval and create independent Scheduled Cron Tasks with dedicated prompts and models.

1. COGNITIVE HEARTBEAT (MAIN SYSTEM LOOP) RUNNING
INTERVAL: 30 SECONDS

Continuously inspects incoming priority task queues (Telegram messages, Alertmanager incidents, VS Code completions). If idle, evaluates active scratchpad context and records health verification telemetry in SQLite.

INSPECTION FREQUENCY: 30s (Default)
TRIAGE REASONING MODEL: CLICK TO EDIT
EVENT DISPATCHER: TOKIO MPSC CHANNEL (100 Cap)
2. USER-SCHEDULED AUTONOMOUS HEARTBEATS {{ heartbeats.length }} REGISTERED JOBS
STATUS ROUTINE NAME CRON EXPR TARGET MODEL MAX TOKENS AUTONOMOUS PROMPT ACTIONS
{{ hb.name }}
{{ hb.cron_expr }}
{{ explainCron(hb.cron_expr) }}
 {{ hb.max_tokens }} {{ hb.prompt }}
No scheduled heartbeats registered. Click "ADD SCHEDULED HEARTBEAT" to automate periodic SRE prompts.

SKILLS & SRE RUNBOOKS

Operational procedures, diagnostic trees, and terminal execution recipes injected into the LLM context.

{{ skill.name }} RUNBOOK

{{ skill.description }}

COMMAND TEMPLATE: {{ skill.command_template || 'None (Reasoning Only)' }}
INSTRUCTIONS EXTRACT: {{ skill.instructions }}
{{ skill.created_at.split(' ')[0] }}
🛠️
No Skills Registered

Register modular skills or runbooks to allow the agent to solve specific infrastructure problems.

TOKEN DIET & SYSTEM CONFIGURATION

Configure global context compression rules, root system prompts, OmniRouter keys, and Governor disjunctor thresholds.

1. OMNIROUTER & INFERENCE GATEWAY
Key is encrypted in SQLite via AES-256-GCM.
Ultra-fast, low-cost model used in background to compress long logs.
Primary reasoning model used for continuous cognitive heartbeat and incoming task triage.
2. GOVERNOR DISJUNCTOR & TOKEN BUDGET
Hard ceiling before triggering automatic semantic summarization.
If confidence decays below this value (default 0.40), repetitive loops are tripped.
Blast Radius Safe-Mode: Require Telegram approval before executing destructive SRE commands
3. CORE AGENT SYSTEM PROMPT ESTIMATED TOKENS: ~{{ (settingsForm.system_prompt.length / 4).toFixed(0) }}

The root persona instructions loaded at the beginning of every autonomous reasoning session.

USER API KEYS & CONNECTORS

Generate Bearer connection keys to plug HAL9k into VS Code (Continue / Roo Code), webhooks, and scripts.

KEY NAME KEY PREFIX SCOPES CREATED AT LAST USED ACTION
{{ key.name }} {{ key.key_prefix }}•••••••• {{ sc }} {{ key.created_at.split(' ')[0] }} {{ key.last_used_at || 'Never' }}
No active API keys found. Generate one to plug HAL9k into VS Code.
INTEGRATION GUIDE: VS CODE (CONTINUE / ROO CODE)

HAL9k exposes an OpenAI-compatible endpoint at /v1/chat/completions. Configure your VS Code extension:

Provider: OpenAI Compatible
Base URL: {{ windowLocationOrigin }}/v1
API Key: hal9k_sk_YOUR_GENERATED_KEY
Model: hal9k-cognitive-v1
{{ getNodeIcon(editingNode.type) }}

{{ editingNode.id }}

{{ editingNode.type }}

Configure execution parameters, model inference, and routing logic.

Supports {{variable}}
{{ editingNode.data.temperature !== undefined ? editingNode.data.temperature : 0.2 }}
Workspace mounted at /workspace
⚡

PIPELINE ACTIVATION & TRIGGER RULES

Define how HAL9k automatically intercepts messages and triggers this pipeline.

HOW ACTIVATION WORKS:

When an incoming message arrives via Telegram, Webhook, or SRE Console, HAL9k checks these rules sequentially. The first matching rule immediately executes this pipeline.

ACTIVATION CONDITIONS ({{ (currentPipelineDef.trigger_conditions || []).length }}):
No custom activation conditions set. Pipeline only runs via explicit /pipeline:{{ currentPipelineDef.id }} or manual execution.
🔍
Core & Recommended
Other Models
⚙️

{{ hoveredModel.id }}

{{ hoveredModel.provider || 'OmniRoute' }} · {{ hoveredModel.id }} · {{ hoveredModel.extended_thinking ? 'extended thinking' : 'standard' }}

Hover a model

Select from the left list to see model specs and context limits.

Max context
{{ hoveredModel.max_context || '128K' }}
{{ hoveredModel.description }}

REGISTER AI PROVIDER

NEW SCHEDULED HEARTBEAT

CREATE SKILL / RUNBOOK

GENERATE API KEY

✓ API KEY GENERATED SUCCESSFULLY

Copy this key now. For security purposes, it will never be displayed in full again.

{{ createdKeyModalData.key }}
WORKSPACE EDITOR: {{ editingFile.path }}
{{ editingFile.content.length }} BYTES

CREATE WORKSPACE FILE

INSTALL MCP SERVER

🧩

CATÁLOGO DE NÓS PARA O DAG

Selecione o tipo de nó para adicionar ao fluxo autônomo.

🧪

TESTAR TOOL LUA // {{ testToolName }}

Plugin: {{ testingPluginTarget ? testingPluginTarget.name : '' }} ({{ testingPluginTarget ? testingPluginTarget.id : '' }})

Validação estrita
Sandbox VM Pronta
{{ testResultOutput.status === 'ok' ? '✓' : '✗' }} {{ testResultOutput.status === 'ok' ? 'Executado com Sucesso' : 'Falha na Execução' }} Latência: {{ testResultOutput.duration_ms }} ms 
{{ testResultOutput.output || testResultOutput.error }}
TASK AUDIT INSPECTOR: {{ selectedTaskDetails.task_code }} {{ selectedTaskDetails.status }}
USER PROMPT:
{{ selectedTaskDetails.prompt }}
EXACT SYSTEM PROMPT USED IN THIS EXECUTION (FROM SQLITE): {{ selectedTaskDetails.system_prompt_used.length }} CHARS
{{ selectedTaskDetails.system_prompt_used }}
📼 TIME-TRAVEL AUDIT REPLAY (STEP {{ replayStepIndex + 1 }} OF {{ selectedTaskDetails.steps.length }}) {{ selectedTaskDetails.steps[replayStepIndex].action }}
Tool Input: {{ JSON.stringify(selectedTaskDetails.steps[replayStepIndex].input) }}
Observation: {{ selectedTaskDetails.steps[replayStepIndex].output }}
AUTONOMOUS STEPS EXECUTED ({{ selectedTaskDetails.steps.length }} STEPS):
STEP #{{ st.step }} — {{ st.action }} CONFIDENCE: {{ (st.confidence * 100).toFixed(0) }}% | {{ st.duration_ms }}ms 
{{ JSON.stringify(st.input, null, 2) }}
{{ st.output }}
FINAL AGENT RESOLUTION:
{{ selectedTaskDetails.final_answer || 'No final answer recorded.' }}
EDIT AGENT TASK: {{ editingTask.task_code }} ID: {{ editingTask.id }}
{{ (editingTask.system_prompt_used || '').length }} CHARS

HAL9k OPERATOR LOGIN

Authenticate via corporate Samba AD DC or local fallback admin account.

{{ toast.type === 'error' ? '⚠️' : '✓' }} {{ toast.message }}