Skip to main content

gbrain

Use when the user asks for GBrain, personal knowledge brain setup, brain-first retrieval, local PGLite brain, hybrid RAG search, gbrain import/query/sync/embed, GBrain skills, Minions, durable agent jobs, soul-audit, brain maintenance, or integrating GBrain with gstack, Hermes, OpenClaw, Claude Code, or Codex.

Zur Installation springen

Quellinformationen

Repository
onfire7777/universal-ai-skills-library
Letzte Quellaktivität
13. Mai 2026 um 23:11
Erkannte Sprache von SKILL.md
Englisch
Sterne
16
Forks
0

Installationsoptionen

Standardmäßig ist der Prompt ausgewählt, der zuerst die Quelle prüft. Sie können zu einem direkten Befehl wechseln oder eine lokale Kopie herunterladen.

Quelldateien prüfen

Lesen Sie SKILL.md und alle von SkillsMP angezeigten Begleitdateien, bevor Sie sich für eine Installation entscheiden.

SKILL.md wird angezeigt

SKILL.md
Quellanweisungen · Schreibgeschützte Vorschau
name
gbrain
description
Use when the user asks for GBrain, personal knowledge brain setup, brain-first retrieval, local PGLite brain, hybrid RAG search, gbrain import/query/sync/embed, GBrain skills, Minions, durable agent jobs, soul-audit, brain maintenance, or integrating GBrain with gstack, Hermes, OpenClaw, Claude Code, or Codex.
# GBrain Universal Adapter Canonical upstream checkout: `%USERPROFILE%\gbrain` GBrain is a real CLI-backed personal knowledge brain. It is not just a skill pack. The universal stack keeps the upstream checkout and local brain state in one place, then exposes GBrain skills and commands through `skill-router`. On Windows, the canonical GBrain binary is `%USERPROFILE%\.bun\bin\gbrain.exe`. The compatibility shim `%USERPROFILE%\go\bin\gbrain.cmd` delegates to it so AI hosts with an older inherited PATH still resolve `gbrain` without duplicating the install. ## Core Commands ```bash gbrain --version gbrain doctor --json gbrain stats gbrain import <dir> --no-embed gbrain embed --stale gbrain search "<query>" gbrain query "<question>" gbrain sync --repo <brain-repo> gbrain jobs smoke ``` Use keyword search without API keys. In the Universal AI Stack, text vector search uses the local Qwen embedding service through `llama-server:qwen3-embedding-0.6b` at `http://127.0.0.1:18084/v1` with 1024 dimensions, so GBrain text embeddings do not require an OpenAI key. Anthropic query expansion and durable subagent workers require the relevant provider key only for those optional features. ## Load GBrain Skills On Demand The upstream GBrain skills are indexed from: ```text %USERPROFILE%\gbrain\skills ``` Examples: ```bash skill-router skill brain-ops skill-router skill signal-detector skill-router skill query skill-router skill ingest skill-router skill media-ingest skill-router skill maintain skill-router skill minion-orchestrator skill-router skill soul-audit skill-router skill skillify ``` Before changing a GBrain workflow, read: ```bash skill-router skill gbrain skill-router skill brain-ops skill-router skill query ``` For setup or repair, read the upstream installer flow first: ```powershell Get-Content "$env:USERPROFILE\gbrain\INSTALL_FOR_AGENTS.md" Get-Content "$env:USERPROFILE\gbrain\AGENTS.md" Get-Content "$env:USERPROFILE\gbrain\docs\GBRAIN_VERIFY.md" ``` ## Integration Policy - Keep `%USERPROFILE%\gbrain` as the single upstream source checkout. - Keep `%USERPROFILE%\.gbrain` for GBrain runtime state such as PGLite data and user-level GBrain skills. - Do not vendor GBrain's full skill tree into every AI root. - Use `skill-router` for on-demand loading and `gbrain` for persistent brain operations. - GStack and GBrain can cooperate: GStack supplies engineering process skills; GBrain supplies brain-first retrieval, memory, jobs, and durable agent context. ## Maintenance Follow the upstream upgrade path: ```powershell git -C "$env:USERPROFILE\gbrain" pull --ff-only Set-Location "$env:USERPROFILE\gbrain" bun install gbrain init gbrain post-upgrade gbrain doctor --json ``` Do not run `bun install -g github:garrytan/gbrain`; upstream documents that this skips required install behavior.
Auf GitHub ansehen