merge: sync upstream/main, wire WebSearch through factory

Merge upstream/main into refactor/provider-protocol-122.
Resolve http_provider.go conflict (keep thin delegate).
Wire OpenAIProviderConfig.WebSearch through providerSelection
and into CodexProvider for codex-auth and codex-cli-token paths.
This commit is contained in:
Leandro Barbosa
2026-02-18 11:09:18 -03:00
23 changed files with 1465 additions and 94 deletions
+18 -12
View File
@@ -1,4 +1,7 @@
## 📝 Description
<!-- Please briefly describe the changes and purpose of this PR -->
## 🗣️ Type of Change
- [ ] 🐞 Bug fix (non-breaking change which fixes an issue)
- [ ] ✨ New feature (non-breaking change which adds functionality)
@@ -11,25 +14,28 @@
- [ ] 👨‍💻 Mostly Human-written (Human lead, AI assisted or none)
## 🔗 Linked Issue
## 🔗 Related Issue
<!-- Please link the related issue(s) (e.g., Fixes #123, Closes #456) -->
## 📚 Technical Context (Skip for Docs)
* **Reference:** [URL]
* **Reasoning:** ...
- **Reference URL:**
- **Reasoning:**
## 🧪 Test Environment
- **Hardware:** <!-- e.g. Raspberry Pi 5, Orange Pi, PC-->
- **OS:** <!-- e.g. Debian 12, Ubuntu 22.04 -->
- **Model/Provider:** <!-- e.g. OpenAI GPT-4o, Kimi k2, DeepSeek-V3 -->
- **Channels:** <!-- e.g. Discord, Telegram, Feishu, ... -->
## 🧪 Test Environment & Hardware
- **Hardware:** [e.g. Raspberry Pi 5, Orange Pi, PC]
- **OS:** [e.g. Debian 12, Ubuntu 22.04]
- **Model/Provider:** [e.g. OpenAI GPT-4o, Kimi k2, DeepSeek-V3]
- **Channels:** [e.g. Discord, Telegram, Feishu, ...]
## 📸 Proof of Work (Optional for Docs)
## 📸 Evidence (Optional)
<details>
<summary>Click to view Logs/Screenshots</summary>
</details>
<!-- Please paste relevant screenshots or logs here -->
</details>
## ☑️ Checklist
- [ ] My code/docs follow the style of this project.
+2
View File
@@ -55,6 +55,7 @@ jobs:
ref: ${{ inputs.tag }}
- name: Setup Go from go.mod
id: setup-go
uses: actions/setup-go@v5
with:
go-version-file: go.mod
@@ -89,6 +90,7 @@ jobs:
GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
GITHUB_REPOSITORY_OWNER: ${{ github.repository_owner }}
DOCKERHUB_IMAGE_NAME: ${{ vars.DOCKERHUB_REPOSITORY }}
GOVERSION: ${{ steps.setup-go.outputs.go-version }}
- name: Apply release flags
shell: bash
+8
View File
@@ -11,6 +11,14 @@ builds:
- id: picoclaw
env:
- CGO_ENABLED=0
tags:
- stdjson
ldflags:
- -s -w
- -X main.version={{ .Version }}
- -X main.gitCommit={{ .ShortCommit }}
- -X main.buildTime={{ .Date }}
- -X main.goVersion={{ .Env.GOVERSION }}
goos:
- linux
- windows
+8 -1
View File
@@ -29,7 +29,14 @@ HEALTHCHECK --interval=30s --timeout=3s --start-period=5s --retries=3 \
# Copy binary
COPY --from=builder /src/build/picoclaw /usr/local/bin/picoclaw
# Create picoclaw home directory
# Create non-root user and group
RUN addgroup -g 1000 picoclaw && \
adduser -D -u 1000 -G picoclaw picoclaw
# Switch to non-root user
USER picoclaw
# Run onboard to create initial directories and config
RUN /usr/local/bin/picoclaw onboard
ENTRYPOINT ["picoclaw"]
+2 -2
View File
@@ -11,11 +11,11 @@ VERSION?=$(shell git describe --tags --always --dirty 2>/dev/null || echo "dev")
GIT_COMMIT=$(shell git rev-parse --short=8 HEAD 2>/dev/null || echo "dev")
BUILD_TIME=$(shell date +%FT%T%z)
GO_VERSION=$(shell $(GO) version | awk '{print $$3}')
LDFLAGS=-ldflags "-X main.version=$(VERSION) -X main.gitCommit=$(GIT_COMMIT) -X main.buildTime=$(BUILD_TIME) -X main.goVersion=$(GO_VERSION)"
LDFLAGS=-ldflags "-X main.version=$(VERSION) -X main.gitCommit=$(GIT_COMMIT) -X main.buildTime=$(BUILD_TIME) -X main.goVersion=$(GO_VERSION) -s -w"
# Go variables
GO?=go
GOFLAGS?=-v
GOFLAGS?=-v -tags stdjson
# Installation
INSTALL_PREFIX?=$(HOME)/.local
+4 -4
View File
@@ -3,7 +3,7 @@
<h1>PicoClaw: Go で書かれた超効率 AI アシスタント</h1>
<h3>$10 ハードウェア · 10MB RAM · 1秒起動 · 皮皮虾,我们走</h3>
<h3>$10 ハードウェア · 10MB RAM · 1秒起動 · 行くぜ、シャコ</h3>
<h3></h3>
<p>
@@ -12,7 +12,7 @@
<img src="https://img.shields.io/badge/license-MIT-green" alt="License">
</p>
**日本語** | [English](README.md)
[中文](README.zh.md) | **日本語** | [Português](README.pt-br.md) | [English](README.md)
</div>
@@ -39,7 +39,7 @@
</table>
## 📢 ニュース
2026-02-09 🎉 PicoClaw リリース!$10 ハードウェアで 10MB 未満の RAM で動く AI エージェントを 1 日で構築。🦐 皮皮虾,我们走
2026-02-09 🎉 PicoClaw リリース!$10 ハードウェアで 10MB 未満の RAM で動く AI エージェントを 1 日で構築。🦐 行くぜ、シャコ
## ✨ 特徴
@@ -735,7 +735,7 @@ Discord: https://discord.gg/V4sAZ9XWpN
## 🐛 トラブルシューティング
### Web 検索で「API 配置问题」と表示される
### Web 検索で「API 設定の問題」と表示される
検索 API キーをまだ設定していない場合、これは正常です。PicoClaw は手動検索用の便利なリンクを提供します。
+1 -1
View File
@@ -14,7 +14,7 @@
<a href="https://x.com/SipeedIO"><img src="https://img.shields.io/badge/X_(Twitter)-SipeedIO-black?style=flat&logo=x&logoColor=white" alt="Twitter"></a>
</p>
[中文](README.zh.md) | [日本語](README.ja.md) | **English**
[中文](README.zh.md) | [日本語](README.ja.md) | [Português](README.pt-br.md) | **English**
</div>
---
+881
View File
@@ -0,0 +1,881 @@
<div align="center">
<img src="assets/logo.jpg" alt="PicoClaw" width="512">
<h1>PicoClaw: Assistente de IA Ultra-Eficiente em Go</h1>
<h3>Hardware de $10 · 10MB de RAM · Boot em 1s · 皮皮虾,我们走!</h3>
<p>
<img src="https://img.shields.io/badge/Go-1.21+-00ADD8?style=flat&logo=go&logoColor=white" alt="Go">
<img src="https://img.shields.io/badge/Arch-x86__64%2C%20ARM64%2C%20RISC--V-blue" alt="Hardware">
<img src="https://img.shields.io/badge/license-MIT-green" alt="License">
<br>
<a href="https://picoclaw.io"><img src="https://img.shields.io/badge/Website-picoclaw.io-blue?style=flat&logo=google-chrome&logoColor=white" alt="Website"></a>
<a href="https://x.com/SipeedIO"><img src="https://img.shields.io/badge/X_(Twitter)-SipeedIO-black?style=flat&logo=x&logoColor=white" alt="Twitter"></a>
</p>
[中文](README.zh.md) | [日本語](README.ja.md) | [English](README.md) | **Português**
</div>
---
🦐 **PicoClaw** é um assistente pessoal de IA ultra-leve inspirado no [nanobot](https://github.com/HKUDS/nanobot), reescrito do zero em **Go** por meio de um processo de "auto-inicialização" (self-bootstrapping) — onde o próprio agente de IA conduziu toda a migração de arquitetura e otimização de código.
⚡️ **Extremamente leve:** Roda em hardware de apenas **$10** com **<10MB** de RAM. Isso é 99% menos memória que o OpenClaw e 98% mais barato que um Mac mini!
<table align="center">
<tr align="center">
<td align="center" valign="top">
<p align="center">
<img src="assets/picoclaw_mem.gif" width="360" height="240">
</p>
</td>
<td align="center" valign="top">
<p align="center">
<img src="assets/licheervnano.png" width="400" height="240">
</p>
</td>
</tr>
</table>
> [!CAUTION]
> **🚨 DECLARACAO DE SEGURANCA & CANAIS OFICIAIS**
>
> * **SEM CRIPTOMOEDAS:** O PicoClaw **NAO** possui nenhum token/moeda oficial. Todas as alegacoes no `pump.fun` ou outras plataformas de negociacao sao **GOLPES**.
> * **DOMINIO OFICIAL:** O **UNICO** site oficial e **[picoclaw.io](https://picoclaw.io)**, e o site da empresa e **[sipeed.com](https://sipeed.com)**.
> * **Aviso:** Muitos dominios `.ai/.org/.com/.net/...` foram registrados por terceiros, nao sao nossos.
> * **Aviso:** O PicoClaw esta em fase inicial de desenvolvimento e pode ter problemas de seguranca de rede nao resolvidos. Nao implante em ambientes de producao antes da versao v1.0.
> * **Nota:** O PicoClaw recentemente fez merge de muitos PRs, o que pode resultar em maior consumo de memoria (10-20MB) nas versoes mais recentes. Planejamos priorizar a otimizacao de recursos assim que o conjunto de funcionalidades estiver estavel.
## 📢 Novidades
2026-02-16 🎉 PicoClaw atingiu 12K stars em uma semana! Obrigado a todos pelo apoio! O PicoClaw esta crescendo mais rapido do que jamais imaginamos. Dado o alto volume de PRs, precisamos urgentemente de maintainers da comunidade. Nossos papeis de voluntarios e roadmap foram publicados oficialmente [aqui](docs/picoclaw_community_roadmap_260216.md) — estamos ansiosos para ter voce a bordo!
2026-02-13 🎉 PicoClaw atingiu 5000 stars em 4 dias! Obrigado a comunidade! Estamos finalizando o **Roadmap do Projeto** e configurando o **Grupo de Desenvolvedores** para acelerar o desenvolvimento do PicoClaw.
🚀 **Chamada para Acao:** Envie suas solicitacoes de funcionalidades nas GitHub Discussions. Revisaremos e priorizaremos na proxima reuniao semanal.
2026-02-09 🎉 PicoClaw lancado oficialmente! Construido em 1 dia para trazer Agentes de IA para hardware de $10 com <10MB de RAM. 🦐 PicoClaw, Partiu!
## ✨ Funcionalidades
🪶 **Ultra-Leve**: Consumo de memoria <10MB — 99% menor que o Clawdbot para funcionalidades essenciais.
💰 **Custo Minimo**: Eficiente o suficiente para rodar em hardware de $10 — 98% mais barato que um Mac mini.
⚡️ **Inicializacao Relampago**: Tempo de inicializacao 400X mais rapido, boot em 1 segundo mesmo em CPU single-core de 0.6GHz.
🌍 **Portabilidade Real**: Um unico binario auto-contido para RISC-V, ARM e x86. Um clique e ja era!
🤖 **Auto-Construido por IA**: Implementacao nativa em Go de forma autonoma — 95% do nucleo gerado pelo Agente com refinamento humano no loop.
| | OpenClaw | NanoBot | **PicoClaw** |
| ----------------------------- | ------------- | ------------------------ | ----------------------------------------- |
| **Linguagem** | TypeScript | Python | **Go** |
| **RAM** | >1GB | >100MB | **< 10MB** |
| **Inicializacao**</br>(CPU 0.8GHz) | >500s | >30s | **<1s** |
| **Custo** | Mac Mini $599 | Maioria dos SBC Linux </br>~$50 | **Qualquer placa Linux**</br>**A partir de $10** |
<img src="assets/compare.jpg" alt="PicoClaw" width="512">
## 🦾 Demonstracao
### 🛠️ Fluxos de Trabalho Padrao do Assistente
<table align="center">
<tr align="center">
<th><p align="center">🧩 Engenharia Full-Stack</p></th>
<th><p align="center">🗂️ Gerenciamento de Logs & Planejamento</p></th>
<th><p align="center">🔎 Busca Web & Aprendizado</p></th>
</tr>
<tr>
<td align="center"><p align="center"><img src="assets/picoclaw_code.gif" width="240" height="180"></p></td>
<td align="center"><p align="center"><img src="assets/picoclaw_memory.gif" width="240" height="180"></p></td>
<td align="center"><p align="center"><img src="assets/picoclaw_search.gif" width="240" height="180"></p></td>
</tr>
<tr>
<td align="center">Desenvolver • Implantar • Escalar</td>
<td align="center">Agendar • Automatizar • Memorizar</td>
<td align="center">Descobrir • Analisar • Tendencias</td>
</tr>
</table>
### 📱 Rode em celulares Android antigos
De uma segunda vida ao seu celular de dez anos atras! Transforme-o em um assistente de IA inteligente com o PicoClaw. Inicio rapido:
1. **Instale o Termux** (Disponivel no F-Droid ou Google Play).
2. **Execute os comandos**
```bash
# Nota: Substitua v0.1.1 pela versao mais recente da pagina de Releases
wget https://github.com/sipeed/picoclaw/releases/download/v0.1.1/picoclaw-linux-arm64
chmod +x picoclaw-linux-arm64
pkg install proot
termux-chroot ./picoclaw-linux-arm64 onboard
```
Depois siga as instrucoes na secao "Inicio Rapido" para completar a configuracao!
<img src="assets/termux.jpg" alt="PicoClaw" width="512">
### 🐜 Implantacao Inovadora com Baixo Consumo
O PicoClaw pode ser implantado em praticamente qualquer dispositivo Linux!
- $9.9 [LicheeRV-Nano](https://www.aliexpress.com/item/1005006519668532.html) versao E (Ethernet) ou W (WiFi6), para Assistente Domestico Minimalista
- $30~50 [NanoKVM](https://www.aliexpress.com/item/1005007369816019.html), ou $100 [NanoKVM-Pro](https://www.aliexpress.com/item/1005010048471263.html) para Manutencao Automatizada de Servidores
- $50 [MaixCAM](https://www.aliexpress.com/item/1005008053333693.html) ou $100 [MaixCAM2](https://www.kickstarter.com/projects/zepan/maixcam2-build-your-next-gen-4k-ai-camera) para Monitoramento Inteligente
https://private-user-images.githubusercontent.com/83055338/547056448-e7b031ff-d6f5-4468-bcca-5726b6fecb5c.mp4
🌟 Mais cenarios de implantacao aguardam voce!
## 📦 Instalacao
### Instalar com binario pre-compilado
Baixe o binario para sua plataforma na pagina de [releases](https://github.com/sipeed/picoclaw/releases).
### Instalar a partir do codigo-fonte (funcionalidades mais recentes, recomendado para desenvolvimento)
```bash
git clone https://github.com/sipeed/picoclaw.git
cd picoclaw
make deps
# Build, sem necessidade de instalar
make build
# Build para multiplas plataformas
make build-all
# Build e Instalar
make install
```
## 🐳 Docker Compose
Voce tambem pode rodar o PicoClaw usando Docker Compose sem instalar nada localmente.
```bash
# 1. Clone este repositorio
git clone https://github.com/sipeed/picoclaw.git
cd picoclaw
# 2. Configure suas API keys
cp config/config.example.json config/config.json
vim config/config.json # Configure DISCORD_BOT_TOKEN, API keys, etc.
# 3. Build & Iniciar
docker compose --profile gateway up -d
# 4. Ver logs
docker compose logs -f picoclaw-gateway
# 5. Parar
docker compose --profile gateway down
```
### Modo Agente (Execucao unica)
```bash
# Fazer uma pergunta
docker compose run --rm picoclaw-agent -m "Quanto e 2+2?"
# Modo interativo
docker compose run --rm picoclaw-agent
```
### Rebuild
```bash
docker compose --profile gateway build --no-cache
docker compose --profile gateway up -d
```
### 🚀 Inicio Rapido
> [!TIP]
> Configure sua API key em `~/.picoclaw/config.json`.
> Obtenha API keys: [OpenRouter](https://openrouter.ai/keys) (LLM) · [Zhipu](https://open.bigmodel.cn/usercenter/proj-mgmt/apikeys) (LLM)
> Busca web e **opcional** — obtenha a [Brave Search API](https://brave.com/search/api) gratuita (2000 consultas gratis/mes) ou use o fallback automatico integrado.
**1. Inicializar**
```bash
picoclaw onboard
```
**2. Configurar** (`~/.picoclaw/config.json`)
```json
{
"agents": {
"defaults": {
"workspace": "~/.picoclaw/workspace",
"model": "glm-4.7",
"max_tokens": 8192,
"temperature": 0.7,
"max_tool_iterations": 20
}
},
"providers": {
"openrouter": {
"api_key": "xxx",
"api_base": "https://openrouter.ai/api/v1"
}
},
"tools": {
"web": {
"brave": {
"enabled": false,
"api_key": "YOUR_BRAVE_API_KEY",
"max_results": 5
},
"duckduckgo": {
"enabled": true,
"max_results": 5
}
}
}
}
```
**3. Obter API Keys**
* **Provedor de LLM**: [OpenRouter](https://openrouter.ai/keys) · [Zhipu](https://open.bigmodel.cn/usercenter/proj-mgmt/apikeys) · [Anthropic](https://console.anthropic.com) · [OpenAI](https://platform.openai.com) · [Gemini](https://aistudio.google.com/api-keys)
* **Busca Web** (opcional): [Brave Search](https://brave.com/search/api) - Plano gratuito disponivel (2000 consultas/mes)
> **Nota**: Veja `config.example.json` para um modelo de configuracao completo.
**4. Conversar**
```bash
picoclaw agent -m "Quanto e 2+2?"
```
Pronto! Voce tem um assistente de IA funcionando em 2 minutos.
---
## 💬 Integracao com Apps de Chat
Converse com seu PicoClaw via Telegram, Discord, DingTalk ou LINE.
| Canal | Nivel de Configuracao |
| --- | --- |
| **Telegram** | Facil (apenas um token) |
| **Discord** | Facil (bot token + intents) |
| **QQ** | Facil (AppID + AppSecret) |
| **DingTalk** | Medio (credenciais do app) |
| **LINE** | Medio (credenciais + webhook URL) |
<details>
<summary><b>Telegram</b> (Recomendado)</summary>
**1. Criar o bot**
* Abra o Telegram, busque `@BotFather`
* Envie `/newbot`, siga as instrucoes
* Copie o token
**2. Configurar**
```json
{
"channels": {
"telegram": {
"enabled": true,
"token": "YOUR_BOT_TOKEN",
"allowFrom": ["YOUR_USER_ID"]
}
}
}
```
> Obtenha seu User ID pelo `@userinfobot` no Telegram.
**3. Executar**
```bash
picoclaw gateway
```
</details>
<details>
<summary><b>Discord</b></summary>
**1. Criar o bot**
* Acesse <https://discord.com/developers/applications>
* Crie um aplicativo → Bot → Add Bot
* Copie o token do bot
**2. Habilitar Intents**
* Nas configuracoes do Bot, habilite **MESSAGE CONTENT INTENT**
* (Opcional) Habilite **SERVER MEMBERS INTENT** se quiser usar lista de permissoes baseada em dados dos membros
**3. Obter seu User ID**
* Configuracoes do Discord → Avancado → habilite **Modo Desenvolvedor**
* Clique com botao direito no seu avatar → **Copiar ID do Usuario**
**4. Configurar**
```json
{
"channels": {
"discord": {
"enabled": true,
"token": "YOUR_BOT_TOKEN",
"allowFrom": ["YOUR_USER_ID"]
}
}
}
```
**5. Convidar o bot**
* OAuth2 → URL Generator
* Scopes: `bot`
* Bot Permissions: `Send Messages`, `Read Message History`
* Abra a URL de convite gerada e adicione o bot ao seu servidor
**6. Executar**
```bash
picoclaw gateway
```
</details>
<details>
<summary><b>QQ</b></summary>
**1. Criar o bot**
- Acesse a [QQ Open Platform](https://q.qq.com/#)
- Crie um aplicativo → Obtenha **AppID** e **AppSecret**
**2. Configurar**
```json
{
"channels": {
"qq": {
"enabled": true,
"app_id": "YOUR_APP_ID",
"app_secret": "YOUR_APP_SECRET",
"allow_from": []
}
}
}
```
> Deixe `allow_from` vazio para permitir todos os usuarios, ou especifique numeros QQ para restringir o acesso.
**3. Executar**
```bash
picoclaw gateway
```
</details>
<details>
<summary><b>DingTalk</b></summary>
**1. Criar o bot**
* Acesse a [Open Platform](https://open.dingtalk.com/)
* Crie um app interno
* Copie o Client ID e Client Secret
**2. Configurar**
```json
{
"channels": {
"dingtalk": {
"enabled": true,
"client_id": "YOUR_CLIENT_ID",
"client_secret": "YOUR_CLIENT_SECRET",
"allow_from": []
}
}
}
```
> Deixe `allow_from` vazio para permitir todos os usuarios, ou especifique IDs para restringir o acesso.
**3. Executar**
```bash
picoclaw gateway
```
</details>
<details>
<summary><b>LINE</b></summary>
**1. Criar uma Conta Oficial LINE**
- Acesse o [LINE Developers Console](https://developers.line.biz/)
- Crie um provider → Crie um canal Messaging API
- Copie o **Channel Secret** e o **Channel Access Token**
**2. Configurar**
```json
{
"channels": {
"line": {
"enabled": true,
"channel_secret": "YOUR_CHANNEL_SECRET",
"channel_access_token": "YOUR_CHANNEL_ACCESS_TOKEN",
"webhook_host": "0.0.0.0",
"webhook_port": 18791,
"webhook_path": "/webhook/line",
"allow_from": []
}
}
}
```
**3. Configurar URL do Webhook**
O LINE requer HTTPS para webhooks. Use um reverse proxy ou tunnel:
```bash
# Exemplo com ngrok
ngrok http 18791
```
Em seguida, configure a Webhook URL no LINE Developers Console para `https://seu-dominio/webhook/line` e habilite **Use webhook**.
**4. Executar**
```bash
picoclaw gateway
```
> Em chats de grupo, o bot responde apenas quando mencionado com @. As respostas citam a mensagem original.
> **Docker Compose**: Adicione `ports: ["18791:18791"]` ao servico `picoclaw-gateway` para expor a porta do webhook.
</details>
## <img src="assets/clawdchat-icon.png" width="24" height="24" alt="ClawdChat"> Junte-se a Rede Social de Agentes
Conecte o PicoClaw a Rede Social de Agentes simplesmente enviando uma unica mensagem via CLI ou qualquer App de Chat integrado.
**Leia `https://clawdchat.ai/skill.md` e siga as instrucoes para entrar no [ClawdChat.ai](https://clawdchat.ai)**
## ⚙️ Configuracao Detalhada
Arquivo de configuracao: `~/.picoclaw/config.json`
### Estrutura do Workspace
O PicoClaw armazena dados no workspace configurado (padrao: `~/.picoclaw/workspace`):
```
~/.picoclaw/workspace/
├── sessions/ # Sessoes de conversa e historico
├── memory/ # Memoria de longo prazo (MEMORY.md)
├── state/ # Estado persistente (ultimo canal, etc.)
├── cron/ # Banco de dados de tarefas agendadas
├── skills/ # Skills personalizadas
├── AGENTS.md # Guia de comportamento do Agente
├── HEARTBEAT.md # Prompts de tarefas periodicas (verificado a cada 30 min)
├── IDENTITY.md # Identidade do Agente
├── SOUL.md # Alma do Agente
├── TOOLS.md # Descricao das ferramentas
└── USER.md # Preferencias do usuario
```
### 🔒 Sandbox de Seguranca
O PicoClaw roda em um ambiente sandbox por padrao. O agente so pode acessar arquivos e executar comandos dentro do workspace configurado.
#### Configuracao Padrao
```json
{
"agents": {
"defaults": {
"workspace": "~/.picoclaw/workspace",
"restrict_to_workspace": true
}
}
}
```
| Opcao | Padrao | Descricao |
|-------|--------|-----------|
| `workspace` | `~/.picoclaw/workspace` | Diretorio de trabalho do agente |
| `restrict_to_workspace` | `true` | Restringir acesso de arquivos/comandos ao workspace |
#### Ferramentas Protegidas
Quando `restrict_to_workspace: true`, as seguintes ferramentas sao restritas ao sandbox:
| Ferramenta | Funcao | Restricao |
|------------|--------|-----------|
| `read_file` | Ler arquivos | Apenas arquivos dentro do workspace |
| `write_file` | Escrever arquivos | Apenas arquivos dentro do workspace |
| `list_dir` | Listar diretorios | Apenas diretorios dentro do workspace |
| `edit_file` | Editar arquivos | Apenas arquivos dentro do workspace |
| `append_file` | Adicionar a arquivos | Apenas arquivos dentro do workspace |
| `exec` | Executar comandos | Caminhos dos comandos devem estar dentro do workspace |
#### Protecao Adicional do Exec
Mesmo com `restrict_to_workspace: false`, a ferramenta `exec` bloqueia estes comandos perigosos:
* `rm -rf`, `del /f`, `rmdir /s` — Exclusao em massa
* `format`, `mkfs`, `diskpart` — Formatacao de disco
* `dd if=` — Criacao de imagem de disco
* Escrita em `/dev/sd[a-z]` — Escrita direta no disco
* `shutdown`, `reboot`, `poweroff` — Desligamento do sistema
* Fork bomb `:(){ :|:& };:`
#### Exemplos de Erro
```
[ERROR] tool: Tool execution failed
{tool=exec, error=Command blocked by safety guard (path outside working dir)}
```
```
[ERROR] tool: Tool execution failed
{tool=exec, error=Command blocked by safety guard (dangerous pattern detected)}
```
#### Desabilitar Restricoes (Risco de Seguranca)
Se voce precisa que o agente acesse caminhos fora do workspace:
**Metodo 1: Arquivo de configuracao**
```json
{
"agents": {
"defaults": {
"restrict_to_workspace": false
}
}
}
```
**Metodo 2: Variavel de ambiente**
```bash
export PICOCLAW_AGENTS_DEFAULTS_RESTRICT_TO_WORKSPACE=false
```
> ⚠️ **Aviso**: Desabilitar esta restricao permite que o agente acesse qualquer caminho no seu sistema. Use com cuidado apenas em ambientes controlados.
#### Consistencia do Limite de Seguranca
A configuracao `restrict_to_workspace` se aplica consistentemente em todos os caminhos de execucao:
| Caminho de Execucao | Limite de Seguranca |
|----------------------|---------------------|
| Agente Principal | `restrict_to_workspace` ✅ |
| Subagente / Spawn | Herda a mesma restricao ✅ |
| Tarefas Heartbeat | Herda a mesma restricao ✅ |
Todos os caminhos compartilham a mesma restricao de workspace — nao ha como contornar o limite de seguranca por meio de subagentes ou tarefas agendadas.
### Heartbeat (Tarefas Periodicas)
O PicoClaw pode executar tarefas periodicas automaticamente. Crie um arquivo `HEARTBEAT.md` no seu workspace:
```markdown
# Tarefas Periodicas
- Verificar meu email para mensagens importantes
- Revisar minha agenda para proximos eventos
- Verificar a previsao do tempo
```
O agente lera este arquivo a cada 30 minutos (configuravel) e executara as tarefas usando as ferramentas disponiveis.
#### Tarefas Assincronas com Spawn
Para tarefas de longa duracao (busca web, chamadas de API), use a ferramenta `spawn` para criar um **subagente**:
```markdown
# Tarefas Periodicas
## Tarefas Rapidas (resposta direta)
- Informar hora atual
## Tarefas Longas (usar spawn para async)
- Buscar noticias de IA na web e resumir
- Verificar email e reportar mensagens importantes
```
**Comportamentos principais:**
| Funcionalidade | Descricao |
|----------------|-----------|
| **spawn** | Cria subagente assincrono, nao bloqueia o heartbeat |
| **Contexto independente** | Subagente tem seu proprio contexto, sem historico de sessao |
| **Ferramenta message** | Subagente se comunica diretamente com o usuario via ferramenta message |
| **Nao-bloqueante** | Apos o spawn, o heartbeat continua para a proxima tarefa |
#### Como Funciona a Comunicacao do Subagente
```
Heartbeat dispara
Agente le HEARTBEAT.md
Para tarefa longa: spawn subagente
↓ ↓
Continua proxima tarefa Subagente trabalha independentemente
↓ ↓
Todas tarefas concluidas Subagente usa ferramenta "message"
↓ ↓
Responde HEARTBEAT_OK Usuario recebe resultado diretamente
```
O subagente tem acesso as ferramentas (message, web_search, etc.) e pode se comunicar com o usuario independentemente sem passar pelo agente principal.
**Configuracao:**
```json
{
"heartbeat": {
"enabled": true,
"interval": 30
}
}
```
| Opcao | Padrao | Descricao |
|-------|--------|-----------|
| `enabled` | `true` | Habilitar/desabilitar heartbeat |
| `interval` | `30` | Intervalo de verificacao em minutos (min: 5) |
**Variaveis de ambiente:**
* `PICOCLAW_HEARTBEAT_ENABLED=false` para desabilitar
* `PICOCLAW_HEARTBEAT_INTERVAL=60` para alterar o intervalo
### Provedores
> [!NOTE]
> O Groq fornece transcricao de voz gratuita via Whisper. Se configurado, mensagens de voz do Telegram serao automaticamente transcritas.
| Provedor | Finalidade | Obter API Key |
| --- | --- | --- |
| `gemini` | LLM (Gemini direto) | [aistudio.google.com](https://aistudio.google.com) |
| `zhipu` | LLM (Zhipu direto) | [bigmodel.cn](bigmodel.cn) |
| `openrouter` (Em teste) | LLM (recomendado, acesso a todos os modelos) | [openrouter.ai](https://openrouter.ai) |
| `anthropic` (Em teste) | LLM (Claude direto) | [console.anthropic.com](https://console.anthropic.com) |
| `openai` (Em teste) | LLM (GPT direto) | [platform.openai.com](https://platform.openai.com) |
| `deepseek` (Em teste) | LLM (DeepSeek direto) | [platform.deepseek.com](https://platform.deepseek.com) |
| `groq` | LLM + **Transcricao de voz** (Whisper) | [console.groq.com](https://console.groq.com) |
<details>
<summary><b>Configuracao Zhipu</b></summary>
**1. Obter API key**
* Obtenha a [API key](https://bigmodel.cn/usercenter/proj-mgmt/apikeys)
**2. Configurar**
```json
{
"agents": {
"defaults": {
"workspace": "~/.picoclaw/workspace",
"model": "glm-4.7",
"max_tokens": 8192,
"temperature": 0.7,
"max_tool_iterations": 20
}
},
"providers": {
"zhipu": {
"api_key": "Sua API Key",
"api_base": "https://open.bigmodel.cn/api/paas/v4"
}
}
}
```
**3. Executar**
```bash
picoclaw agent -m "Ola, como vai?"
```
</details>
<details>
<summary><b>Exemplo de configuracao completa</b></summary>
```json
{
"agents": {
"defaults": {
"model": "anthropic/claude-opus-4-5"
}
},
"providers": {
"openrouter": {
"api_key": "sk-or-v1-xxx"
},
"groq": {
"api_key": "gsk_xxx"
}
},
"channels": {
"telegram": {
"enabled": true,
"token": "123456:ABC...",
"allow_from": ["123456789"]
},
"discord": {
"enabled": true,
"token": "",
"allow_from": [""]
},
"whatsapp": {
"enabled": false
},
"feishu": {
"enabled": false,
"app_id": "cli_xxx",
"app_secret": "xxx",
"encrypt_key": "",
"verification_token": "",
"allow_from": []
},
"qq": {
"enabled": false,
"app_id": "",
"app_secret": "",
"allow_from": []
}
},
"tools": {
"web": {
"brave": {
"enabled": false,
"api_key": "BSA...",
"max_results": 5
},
"duckduckgo": {
"enabled": true,
"max_results": 5
}
},
"cron": {
"exec_timeout_minutes": 5
}
},
"heartbeat": {
"enabled": true,
"interval": 30
}
}
```
</details>
## Referencia CLI
| Comando | Descricao |
| --- | --- |
| `picoclaw onboard` | Inicializar configuracao & workspace |
| `picoclaw agent -m "..."` | Conversar com o agente |
| `picoclaw agent` | Modo de chat interativo |
| `picoclaw gateway` | Iniciar o gateway (para bots de chat) |
| `picoclaw status` | Mostrar status |
| `picoclaw cron list` | Listar todas as tarefas agendadas |
| `picoclaw cron add ...` | Adicionar uma tarefa agendada |
### Tarefas Agendadas / Lembretes
O PicoClaw suporta lembretes agendados e tarefas recorrentes por meio da ferramenta `cron`:
* **Lembretes unicos**: "Remind me in 10 minutes" (Me lembre em 10 minutos) → dispara uma vez apos 10min
* **Tarefas recorrentes**: "Remind me every 2 hours" (Me lembre a cada 2 horas) → dispara a cada 2 horas
* **Expressoes Cron**: "Remind me at 9am daily" (Me lembre as 9h todos os dias) → usa expressao cron
As tarefas sao armazenadas em `~/.picoclaw/workspace/cron/` e processadas automaticamente.
## 🤝 Contribuir & Roadmap
PRs sao bem-vindos! O codigo-fonte e intencionalmente pequeno e legivel. 🤗
Roadmap em breve...
Grupo de desenvolvedores em formacao. Requisito de entrada: Pelo menos 1 PR com merge.
Grupos de usuarios:
Discord: <https://discord.gg/V4sAZ9XWpN>
<img src="assets/wechat.png" alt="PicoClaw" width="512">
## 🐛 Solucao de Problemas
### Busca web mostra "API 配置问题"
Isso e normal se voce ainda nao configurou uma API key de busca. O PicoClaw fornecera links uteis para busca manual.
Para habilitar a busca web:
1. **Opcao 1 (Recomendado)**: Obtenha uma API key gratuita em [https://brave.com/search/api](https://brave.com/search/api) (2000 consultas gratis/mes) para os melhores resultados.
2. **Opcao 2 (Sem Cartao de Credito)**: Se voce nao tem uma key, o sistema automaticamente usa o **DuckDuckGo** como fallback (sem necessidade de key).
Adicione a key em `~/.picoclaw/config.json` se usar o Brave:
```json
{
"tools": {
"web": {
"brave": {
"enabled": true,
"api_key": "YOUR_BRAVE_API_KEY",
"max_results": 5
},
"duckduckgo": {
"enabled": true,
"max_results": 5
}
}
}
}
```
### Erros de filtragem de conteudo
Alguns provedores (como Zhipu) possuem filtragem de conteudo. Tente reformular sua pergunta ou use um modelo diferente.
### Bot do Telegram diz "Conflict: terminated by other getUpdates"
Isso acontece quando outra instancia do bot esta rodando. Certifique-se de que apenas um `picoclaw gateway` esteja rodando por vez.
---
## 📝 Comparacao de API Keys
| Servico | Plano Gratuito | Caso de Uso |
| --- | --- | --- |
| **OpenRouter** | 200K tokens/mes | Multiplos modelos (Claude, GPT-4, etc.) |
| **Zhipu** | 200K tokens/mes | Melhor para usuarios chineses |
| **Brave Search** | 2000 consultas/mes | Funcionalidade de busca web |
| **Groq** | Plano gratuito disponivel | Inferencia ultra-rapida (Llama, Mixtral) |
+1 -1
View File
@@ -14,7 +14,7 @@
<a href="https://x.com/SipeedIO"><img src="https://img.shields.io/badge/X_(Twitter)-SipeedIO-black?style=flat&logo=x&logoColor=white" alt="Twitter"></a>
</p>
**中文** | [日本語](README.ja.md) | [English](README.md)
**中文** | [日本語](README.ja.md) | [Português](README.pt-br.md) | [English](README.md)
</div>
---
+3 -3
View File
@@ -563,7 +563,7 @@ func gatewayCmd() {
// Setup cron tool and service
execTimeout := time.Duration(cfg.Tools.Cron.ExecTimeoutMinutes) * time.Minute
cronService := setupCronTool(agentLoop, msgBus, cfg.WorkspacePath(), cfg.Agents.Defaults.RestrictToWorkspace, execTimeout)
cronService := setupCronTool(agentLoop, msgBus, cfg.WorkspacePath(), cfg.Agents.Defaults.RestrictToWorkspace, execTimeout, cfg)
heartbeatService := heartbeat.NewHeartbeatService(
cfg.WorkspacePath(),
@@ -988,14 +988,14 @@ func getConfigPath() string {
return filepath.Join(home, ".picoclaw", "config.json")
}
func setupCronTool(agentLoop *agent.AgentLoop, msgBus *bus.MessageBus, workspace string, restrict bool, execTimeout time.Duration) *cron.CronService {
func setupCronTool(agentLoop *agent.AgentLoop, msgBus *bus.MessageBus, workspace string, restrict bool, execTimeout time.Duration, config *config.Config) *cron.CronService {
cronStorePath := filepath.Join(workspace, "cron", "jobs.json")
// Create cron service
cronService := cron.NewCronService(cronStorePath, nil)
// Create and register CronTool
cronTool := tools.NewCronTool(cronService, agentLoop, msgBus, workspace, restrict, execTimeout)
cronTool := tools.NewCronTool(cronService, agentLoop, msgBus, workspace, restrict, execTimeout, config)
agentLoop.RegisterTool(cronTool)
// Set the onJob handler
+3 -2
View File
@@ -79,7 +79,8 @@
},
"openai": {
"api_key": "",
"api_base": ""
"api_base": "",
"web_search": true
},
"openrouter": {
"api_key": "sk-or-v1-xxx",
@@ -144,4 +145,4 @@
"host": "0.0.0.0",
"port": 18790
}
}
}
+4 -4
View File
@@ -11,8 +11,8 @@ services:
profiles:
- agent
volumes:
- ./config/config.json:/root/.picoclaw/config.json:ro
- picoclaw-workspace:/root/.picoclaw/workspace
- ./config/config.json:/home/picoclaw/.picoclaw/config.json:ro
- picoclaw-workspace:/home/picoclaw/.picoclaw/workspace
entrypoint: ["picoclaw", "agent"]
stdin_open: true
tty: true
@@ -31,9 +31,9 @@ services:
- gateway
volumes:
# Configuration file
- ./config/config.json:/root/.picoclaw/config.json:ro
- ./config/config.json:/home/picoclaw/.picoclaw/config.json:ro
# Persistent workspace (sessions, memory, logs)
- picoclaw-workspace:/root/.picoclaw/workspace
- picoclaw-workspace:/home/picoclaw/.picoclaw/workspace
command: ["gateway"]
volumes:
+122
View File
@@ -0,0 +1,122 @@
# Tools Configuration
PicoClaw's tools configuration is located in the `tools` field of `config.json`.
## Directory Structure
```json
{
"tools": {
"web": { ... },
"exec": { ... },
"approval": { ... },
"cron": { ... }
}
}
```
## Web Tools
Web tools are used for web search and fetching.
### Brave
| Config | Type | Default | Description |
|--------|------|---------|-------------|
| `enabled` | bool | false | Enable Brave search |
| `api_key` | string | - | Brave Search API key |
| `max_results` | int | 5 | Maximum number of results |
### DuckDuckGo
| Config | Type | Default | Description |
|--------|------|---------|-------------|
| `enabled` | bool | true | Enable DuckDuckGo search |
| `max_results` | int | 5 | Maximum number of results |
### Perplexity
| Config | Type | Default | Description |
|--------|------|---------|-------------|
| `enabled` | bool | false | Enable Perplexity search |
| `api_key` | string | - | Perplexity API key |
| `max_results` | int | 5 | Maximum number of results |
## Exec Tool
The exec tool is used to execute shell commands.
| Config | Type | Default | Description |
|--------|------|---------|-------------|
| `enable_deny_patterns` | bool | true | Enable default dangerous command blocking |
| `custom_deny_patterns` | array | [] | Custom deny patterns (regular expressions) |
### Functionality
- **`enable_deny_patterns`**: Set to `false` to completely disable the default dangerous command blocking patterns
- **`custom_deny_patterns`**: Add custom deny regex patterns; commands matching these will be blocked
### Default Blocked Command Patterns
By default, PicoClaw blocks the following dangerous commands:
- Delete commands: `rm -rf`, `del /f/q`, `rmdir /s`
- Disk operations: `format`, `mkfs`, `diskpart`, `dd if=`, writing to `/dev/sd*`
- System operations: `shutdown`, `reboot`, `poweroff`
- Command substitution: `$()`, `${}`, backticks
- Pipe to shell: `| sh`, `| bash`
- Privilege escalation: `sudo`, `chmod`, `chown`
- Process control: `pkill`, `killall`, `kill -9`
- Remote operations: `curl | sh`, `wget | sh`, `ssh`
- Package management: `apt`, `yum`, `dnf`, `npm install -g`, `pip install --user`
- Containers: `docker run`, `docker exec`
- Git: `git push`, `git force`
- Other: `eval`, `source *.sh`
### Configuration Example
```json
{
"tools": {
"exec": {
"enable_deny_patterns": true,
"custom_deny_patterns": [
"\\brm\\s+-r\\b",
"\\bkillall\\s+python"
],
}
}
}
```
## Approval Tool
The approval tool controls permissions for dangerous operations.
| Config | Type | Default | Description |
|--------|------|---------|-------------|
| `enabled` | bool | true | Enable approval functionality |
| `write_file` | bool | true | Require approval for file writes |
| `edit_file` | bool | true | Require approval for file edits |
| `append_file` | bool | true | Require approval for file appends |
| `exec` | bool | true | Require approval for command execution |
| `timeout_minutes` | int | 5 | Approval timeout in minutes |
## Cron Tool
The cron tool is used for scheduling periodic tasks.
| Config | Type | Default | Description |
|--------|------|---------|-------------|
| `exec_timeout_minutes` | int | 5 | Execution timeout in minutes, 0 means no limit |
## Environment Variables
All configuration options can be overridden via environment variables with the format `PICOCLAW_TOOLS_<SECTION>_<KEY>`:
For example:
- `PICOCLAW_TOOLS_WEB_BRAVE_ENABLED=true`
- `PICOCLAW_TOOLS_EXEC_ENABLE_DENY_PATTERNS=false`
- `PICOCLAW_TOOLS_CRON_EXEC_TIMEOUT_MINUTES=10`
Note: Array-type environment variables are not currently supported and must be set via the config file.
+1 -1
View File
@@ -71,7 +71,7 @@ func createToolRegistry(workspace string, restrict bool, cfg *config.Config, msg
registry.Register(tools.NewAppendFileTool(workspace, restrict))
// Shell execution
registry.Register(tools.NewExecTool(workspace, restrict))
registry.Register(tools.NewExecToolWithConfig(workspace, restrict, cfg))
if searchTool := tools.NewWebSearchTool(tools.WebSearchToolOptions{
BraveAPIKey: cfg.Tools.Web.Brave.APIKey,
-2
View File
@@ -18,7 +18,6 @@ type MaixCamChannel struct {
listener net.Listener
clients map[net.Conn]bool
clientsMux sync.RWMutex
running bool
}
type MaixCamMessage struct {
@@ -35,7 +34,6 @@ func NewMaixCamChannel(cfg config.MaixCamConfig, bus *bus.MessageBus) (*MaixCamC
BaseChannel: base,
config: cfg,
clients: make(map[net.Conn]bool),
running: false,
}, nil
}
+28 -14
View File
@@ -167,19 +167,19 @@ type DevicesConfig struct {
}
type ProvidersConfig struct {
Anthropic ProviderConfig `json:"anthropic"`
OpenAI ProviderConfig `json:"openai"`
OpenRouter ProviderConfig `json:"openrouter"`
Groq ProviderConfig `json:"groq"`
Zhipu ProviderConfig `json:"zhipu"`
VLLM ProviderConfig `json:"vllm"`
Gemini ProviderConfig `json:"gemini"`
Nvidia ProviderConfig `json:"nvidia"`
Ollama ProviderConfig `json:"ollama"`
Moonshot ProviderConfig `json:"moonshot"`
ShengSuanYun ProviderConfig `json:"shengsuanyun"`
DeepSeek ProviderConfig `json:"deepseek"`
GitHubCopilot ProviderConfig `json:"github_copilot"`
Anthropic ProviderConfig `json:"anthropic"`
OpenAI OpenAIProviderConfig `json:"openai"`
OpenRouter ProviderConfig `json:"openrouter"`
Groq ProviderConfig `json:"groq"`
Zhipu ProviderConfig `json:"zhipu"`
VLLM ProviderConfig `json:"vllm"`
Gemini ProviderConfig `json:"gemini"`
Nvidia ProviderConfig `json:"nvidia"`
Ollama ProviderConfig `json:"ollama"`
Moonshot ProviderConfig `json:"moonshot"`
ShengSuanYun ProviderConfig `json:"shengsuanyun"`
DeepSeek ProviderConfig `json:"deepseek"`
GitHubCopilot ProviderConfig `json:"github_copilot"`
}
type ProviderConfig struct {
@@ -190,6 +190,11 @@ type ProviderConfig struct {
ConnectMode string `json:"connect_mode,omitempty" env:"PICOCLAW_PROVIDERS_{{.Name}}_CONNECT_MODE"` //only for Github Copilot, `stdio` or `grpc`
}
type OpenAIProviderConfig struct {
ProviderConfig
WebSearch bool `json:"web_search" env:"PICOCLAW_PROVIDERS_OPENAI_WEB_SEARCH"`
}
type GatewayConfig struct {
Host string `json:"host" env:"PICOCLAW_GATEWAY_HOST"`
Port int `json:"port" env:"PICOCLAW_GATEWAY_PORT"`
@@ -222,9 +227,15 @@ type CronToolsConfig struct {
ExecTimeoutMinutes int `json:"exec_timeout_minutes" env:"PICOCLAW_TOOLS_CRON_EXEC_TIMEOUT_MINUTES"` // 0 means no timeout
}
type ExecConfig struct {
EnableDenyPatterns bool `json:"enable_deny_patterns" env:"PICOCLAW_TOOLS_EXEC_ENABLE_DENY_PATTERNS"`
CustomDenyPatterns []string `json:"custom_deny_patterns" env:"PICOCLAW_TOOLS_EXEC_CUSTOM_DENY_PATTERNS"`
}
type ToolsConfig struct {
Web WebToolsConfig `json:"web"`
Cron CronToolsConfig `json:"cron"`
Exec ExecConfig `json:"exec"`
}
func DefaultConfig() *Config {
@@ -308,7 +319,7 @@ func DefaultConfig() *Config {
},
Providers: ProvidersConfig{
Anthropic: ProviderConfig{},
OpenAI: ProviderConfig{},
OpenAI: OpenAIProviderConfig{WebSearch: true},
OpenRouter: ProviderConfig{},
Groq: ProviderConfig{},
Zhipu: ProviderConfig{},
@@ -342,6 +353,9 @@ func DefaultConfig() *Config {
Cron: CronToolsConfig{
ExecTimeoutMinutes: 5, // default 5 minutes for LLM operations
},
Exec: ExecConfig{
EnableDenyPatterns: true,
},
},
Heartbeat: HeartbeatConfig{
Enabled: true,
+39
View File
@@ -204,3 +204,42 @@ func TestConfig_Complete(t *testing.T) {
t.Error("Heartbeat should be enabled by default")
}
}
func TestDefaultConfig_OpenAIWebSearchEnabled(t *testing.T) {
cfg := DefaultConfig()
if !cfg.Providers.OpenAI.WebSearch {
t.Fatal("DefaultConfig().Providers.OpenAI.WebSearch should be true")
}
}
func TestLoadConfig_OpenAIWebSearchDefaultsTrueWhenUnset(t *testing.T) {
dir := t.TempDir()
configPath := filepath.Join(dir, "config.json")
if err := os.WriteFile(configPath, []byte(`{"providers":{"openai":{"api_base":""}}}`), 0o600); err != nil {
t.Fatalf("WriteFile() error: %v", err)
}
cfg, err := LoadConfig(configPath)
if err != nil {
t.Fatalf("LoadConfig() error: %v", err)
}
if !cfg.Providers.OpenAI.WebSearch {
t.Fatal("OpenAI codex web search should remain true when unset in config file")
}
}
func TestLoadConfig_OpenAIWebSearchCanBeDisabled(t *testing.T) {
dir := t.TempDir()
configPath := filepath.Join(dir, "config.json")
if err := os.WriteFile(configPath, []byte(`{"providers":{"openai":{"web_search":false}}}`), 0o600); err != nil {
t.Fatalf("WriteFile() error: %v", err)
}
cfg, err := LoadConfig(configPath)
if err != nil {
t.Fatalf("LoadConfig() error: %v", err)
}
if cfg.Providers.OpenAI.WebSearch {
t.Fatal("OpenAI codex web search should be false when disabled in config file")
}
}
+11 -1
View File
@@ -108,7 +108,10 @@ func ConvertConfig(data map[string]interface{}) (*config.Config, []string, error
case "anthropic":
cfg.Providers.Anthropic = pc
case "openai":
cfg.Providers.OpenAI = pc
cfg.Providers.OpenAI = config.OpenAIProviderConfig{
ProviderConfig: pc,
WebSearch: getBoolOrDefault(pMap, "web_search", true),
}
case "openrouter":
cfg.Providers.OpenRouter = pc
case "groq":
@@ -363,6 +366,13 @@ func getBool(data map[string]interface{}, key string) (bool, bool) {
return b, ok
}
func getBoolOrDefault(data map[string]interface{}, key string, defaultVal bool) bool {
if v, ok := getBool(data, key); ok {
return v
}
return defaultVal
}
func getStringSlice(data map[string]interface{}, key string) []string {
v, ok := data[key]
if !ok {
+59 -18
View File
@@ -18,9 +18,10 @@ const codexDefaultModel = "gpt-5.2"
const codexDefaultInstructions = "You are Codex, a coding assistant."
type CodexProvider struct {
client *openai.Client
accountID string
tokenSource func() (string, string, error)
client *openai.Client
accountID string
tokenSource func() (string, string, error)
enableWebSearch bool
}
const defaultCodexInstructions = "You are Codex, a coding assistant."
@@ -37,8 +38,9 @@ func NewCodexProvider(token, accountID string) *CodexProvider {
}
client := openai.NewClient(opts...)
return &CodexProvider{
client: &client,
accountID: accountID,
client: &client,
accountID: accountID,
enableWebSearch: true,
}
}
@@ -78,7 +80,7 @@ func (p *CodexProvider) Chat(ctx context.Context, messages []Message, tools []To
})
}
params := buildCodexParams(messages, tools, resolvedModel, options)
params := buildCodexParams(messages, tools, resolvedModel, options, p.enableWebSearch)
stream := p.client.Responses.NewStreaming(ctx, params, opts...)
defer stream.Close()
@@ -182,7 +184,7 @@ func resolveCodexModel(model string) (string, string) {
return codexDefaultModel, "unsupported model family"
}
func buildCodexParams(messages []Message, tools []ToolDefinition, model string, options map[string]interface{}) responses.ResponseNewParams {
func buildCodexParams(messages []Message, tools []ToolDefinition, model string, options map[string]interface{}, enableWebSearch bool) responses.ResponseNewParams {
var inputItems responses.ResponseInputParam
var instructions string
@@ -217,12 +219,18 @@ func buildCodexParams(messages []Message, tools []ToolDefinition, model string,
})
}
for _, tc := range msg.ToolCalls {
argsJSON, _ := json.Marshal(tc.Arguments)
name, args, ok := resolveCodexToolCall(tc)
if !ok {
logger.WarnCF("provider.codex", "Skipping invalid tool call in history", map[string]interface{}{
"call_id": tc.ID,
})
continue
}
inputItems = append(inputItems, responses.ResponseInputItemUnionParam{
OfFunctionCall: &responses.ResponseFunctionToolCallParam{
CallID: tc.ID,
Name: tc.Name,
Arguments: string(argsJSON),
Name: name,
Arguments: args,
},
})
}
@@ -260,20 +268,50 @@ func buildCodexParams(messages []Message, tools []ToolDefinition, model string,
params.Instructions = openai.Opt(defaultCodexInstructions)
}
if maxTokens, ok := options["max_tokens"].(int); ok {
params.MaxOutputTokens = openai.Opt(int64(maxTokens))
}
if len(tools) > 0 {
params.Tools = translateToolsForCodex(tools)
if len(tools) > 0 || enableWebSearch {
params.Tools = translateToolsForCodex(tools, enableWebSearch)
}
return params
}
func translateToolsForCodex(tools []ToolDefinition) []responses.ToolUnionParam {
result := make([]responses.ToolUnionParam, 0, len(tools))
func resolveCodexToolCall(tc ToolCall) (name string, arguments string, ok bool) {
name = tc.Name
if name == "" && tc.Function != nil {
name = tc.Function.Name
}
if name == "" {
return "", "", false
}
if len(tc.Arguments) > 0 {
argsJSON, err := json.Marshal(tc.Arguments)
if err != nil {
return "", "", false
}
return name, string(argsJSON), true
}
if tc.Function != nil && tc.Function.Arguments != "" {
return name, tc.Function.Arguments, true
}
return name, "{}", true
}
func translateToolsForCodex(tools []ToolDefinition, enableWebSearch bool) []responses.ToolUnionParam {
capHint := len(tools)
if enableWebSearch {
capHint++
}
result := make([]responses.ToolUnionParam, 0, capHint)
for _, t := range tools {
if t.Type != "function" {
continue
}
if enableWebSearch && strings.EqualFold(t.Function.Name, "web_search") {
continue
}
ft := responses.FunctionToolParam{
Name: t.Function.Name,
Parameters: t.Function.Parameters,
@@ -284,6 +322,9 @@ func translateToolsForCodex(tools []ToolDefinition) []responses.ToolUnionParam {
}
result = append(result, responses.ToolUnionParam{OfFunction: &ft})
}
if enableWebSearch {
result = append(result, responses.ToolParamOfWebSearch(responses.WebSearchToolTypeWebSearch))
}
return result
}
+172 -5
View File
@@ -19,7 +19,7 @@ func TestBuildCodexParams_BasicMessage(t *testing.T) {
params := buildCodexParams(messages, nil, "gpt-4o", map[string]interface{}{
"max_tokens": 2048,
"temperature": 0.7,
})
}, true)
if params.Model != "gpt-4o" {
t.Errorf("Model = %q, want %q", params.Model, "gpt-4o")
}
@@ -29,6 +29,9 @@ func TestBuildCodexParams_BasicMessage(t *testing.T) {
if params.Instructions.Or("") != defaultCodexInstructions {
t.Errorf("Instructions = %q, want %q", params.Instructions.Or(""), defaultCodexInstructions)
}
if params.MaxOutputTokens.Valid() {
t.Fatalf("MaxOutputTokens should not be set for Codex backend")
}
}
func TestBuildCodexParams_SystemAsInstructions(t *testing.T) {
@@ -36,7 +39,7 @@ func TestBuildCodexParams_SystemAsInstructions(t *testing.T) {
{Role: "system", Content: "You are helpful"},
{Role: "user", Content: "Hi"},
}
params := buildCodexParams(messages, nil, "gpt-4o", map[string]interface{}{})
params := buildCodexParams(messages, nil, "gpt-4o", map[string]interface{}{}, true)
if !params.Instructions.Valid() {
t.Fatal("Instructions should be set")
}
@@ -56,7 +59,7 @@ func TestBuildCodexParams_ToolCallConversation(t *testing.T) {
},
{Role: "tool", Content: `{"temp": 72}`, ToolCallID: "call_1"},
}
params := buildCodexParams(messages, nil, "gpt-4o", map[string]interface{}{})
params := buildCodexParams(messages, nil, "gpt-4o", map[string]interface{}{}, false)
if params.Input.OfInputItemList == nil {
t.Fatal("Input.OfInputItemList should not be nil")
}
@@ -65,6 +68,45 @@ func TestBuildCodexParams_ToolCallConversation(t *testing.T) {
}
}
func TestBuildCodexParams_ToolCallFunctionFallback(t *testing.T) {
messages := []Message{
{Role: "user", Content: "Read a file"},
{
Role: "assistant",
ToolCalls: []ToolCall{
{
ID: "call_1",
Type: "function",
Function: &FunctionCall{
Name: "read_file",
Arguments: `{"path":"README.md"}`,
},
},
},
},
{Role: "tool", Content: "ok", ToolCallID: "call_1"},
}
params := buildCodexParams(messages, nil, "gpt-4o", map[string]interface{}{}, false)
if params.Input.OfInputItemList == nil {
t.Fatal("Input.OfInputItemList should not be nil")
}
if len(params.Input.OfInputItemList) != 3 {
t.Fatalf("len(Input items) = %d, want 3", len(params.Input.OfInputItemList))
}
fc := params.Input.OfInputItemList[1].OfFunctionCall
if fc == nil {
t.Fatal("assistant tool call should be converted to function_call input item")
}
if fc.Name != "read_file" {
t.Errorf("Function call name = %q, want %q", fc.Name, "read_file")
}
if fc.Arguments != `{"path":"README.md"}` {
t.Errorf("Function call arguments = %q, want %q", fc.Arguments, `{"path":"README.md"}`)
}
}
func TestBuildCodexParams_WithTools(t *testing.T) {
tools := []ToolDefinition{
{
@@ -81,7 +123,7 @@ func TestBuildCodexParams_WithTools(t *testing.T) {
},
},
}
params := buildCodexParams([]Message{{Role: "user", Content: "Hi"}}, tools, "gpt-4o", map[string]interface{}{})
params := buildCodexParams([]Message{{Role: "user", Content: "Hi"}}, tools, "gpt-4o", map[string]interface{}{}, false)
if len(params.Tools) != 1 {
t.Fatalf("len(Tools) = %d, want 1", len(params.Tools))
}
@@ -94,12 +136,61 @@ func TestBuildCodexParams_WithTools(t *testing.T) {
}
func TestBuildCodexParams_StoreIsFalse(t *testing.T) {
params := buildCodexParams([]Message{{Role: "user", Content: "Hi"}}, nil, "gpt-4o", map[string]interface{}{})
params := buildCodexParams([]Message{{Role: "user", Content: "Hi"}}, nil, "gpt-4o", map[string]interface{}{}, false)
if !params.Store.Valid() || params.Store.Or(true) != false {
t.Error("Store should be explicitly set to false")
}
}
func TestBuildCodexParams_DefaultWebSearchEnabled(t *testing.T) {
params := buildCodexParams([]Message{{Role: "user", Content: "Hi"}}, nil, "gpt-4o", map[string]interface{}{}, true)
if len(params.Tools) != 1 {
t.Fatalf("len(Tools) = %d, want 1", len(params.Tools))
}
if params.Tools[0].OfWebSearch == nil {
t.Fatal("Tool should include built-in web_search")
}
if params.Tools[0].OfWebSearch.Type != responses.WebSearchToolTypeWebSearch {
t.Errorf("Web search tool type = %q, want %q", params.Tools[0].OfWebSearch.Type, responses.WebSearchToolTypeWebSearch)
}
}
func TestBuildCodexParams_WebSearchFunctionReplacedWithBuiltin(t *testing.T) {
tools := []ToolDefinition{
{
Type: "function",
Function: ToolFunctionDefinition{
Name: "web_search",
Description: "local web search",
Parameters: map[string]interface{}{
"type": "object",
},
},
},
{
Type: "function",
Function: ToolFunctionDefinition{
Name: "read_file",
Description: "read file",
Parameters: map[string]interface{}{
"type": "object",
},
},
},
}
params := buildCodexParams([]Message{{Role: "user", Content: "Hi"}}, tools, "gpt-4o", map[string]interface{}{}, true)
if len(params.Tools) != 2 {
t.Fatalf("len(Tools) = %d, want 2", len(params.Tools))
}
if params.Tools[0].OfFunction == nil || params.Tools[0].OfFunction.Name != "read_file" {
t.Fatalf("first tool should be function read_file, got %#v", params.Tools[0])
}
if params.Tools[1].OfWebSearch == nil {
t.Fatalf("second tool should be built-in web_search, got %#v", params.Tools[1])
}
}
func TestParseCodexResponse_TextOutput(t *testing.T) {
respJSON := `{
"id": "resp_test",
@@ -214,6 +305,20 @@ func TestCodexProvider_ChatRoundTrip(t *testing.T) {
http.Error(w, "stream must be true", http.StatusBadRequest)
return
}
if _, ok := reqBody["max_output_tokens"]; ok {
http.Error(w, "max_output_tokens is not supported", http.StatusBadRequest)
return
}
toolsAny, ok := reqBody["tools"].([]interface{})
if !ok || len(toolsAny) != 1 {
http.Error(w, "missing default web search tool", http.StatusBadRequest)
return
}
toolObj, ok := toolsAny[0].(map[string]interface{})
if !ok || toolObj["type"] != "web_search" {
http.Error(w, "expected web_search tool", http.StatusBadRequest)
return
}
resp := map[string]interface{}{
"id": "resp_test",
@@ -261,6 +366,64 @@ func TestCodexProvider_ChatRoundTrip(t *testing.T) {
}
}
func TestCodexProvider_ChatRoundTrip_WebSearchDisabled(t *testing.T) {
server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
if r.URL.Path != "/responses" {
http.Error(w, "not found: "+r.URL.Path, http.StatusNotFound)
return
}
var reqBody map[string]interface{}
if err := json.NewDecoder(r.Body).Decode(&reqBody); err != nil {
http.Error(w, "invalid json", http.StatusBadRequest)
return
}
if _, ok := reqBody["tools"]; ok {
http.Error(w, "tools should be absent when web search disabled", http.StatusBadRequest)
return
}
resp := map[string]interface{}{
"id": "resp_test",
"object": "response",
"status": "completed",
"output": []map[string]interface{}{
{
"id": "msg_1",
"type": "message",
"role": "assistant",
"status": "completed",
"content": []map[string]interface{}{
{"type": "output_text", "text": "Hi from Codex!"},
},
},
},
"usage": map[string]interface{}{
"input_tokens": 4,
"output_tokens": 3,
"total_tokens": 7,
"input_tokens_details": map[string]interface{}{"cached_tokens": 0},
"output_tokens_details": map[string]interface{}{"reasoning_tokens": 0},
},
}
writeCompletedSSE(w, resp)
}))
defer server.Close()
provider := NewCodexProvider("test-token", "acc-123")
provider.enableWebSearch = false
provider.client = createOpenAITestClient(server.URL, "test-token", "acc-123")
messages := []Message{{Role: "user", Content: "Hello"}}
resp, err := provider.Chat(t.Context(), messages, nil, "gpt-4o", map[string]interface{}{})
if err != nil {
t.Fatalf("Chat() error: %v", err)
}
if resp.Content != "Hi from Codex!" {
t.Errorf("Content = %q, want %q", resp.Content, "Hi from Codex!")
}
}
func TestCodexProvider_ChatRoundTrip_TokenSourceFallbackAccountID(t *testing.T) {
server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
if r.URL.Path != "/responses" {
@@ -293,6 +456,10 @@ func TestCodexProvider_ChatRoundTrip_TokenSourceFallbackAccountID(t *testing.T)
http.Error(w, "temperature is not supported", http.StatusBadRequest)
return
}
if _, ok := reqBody["max_output_tokens"]; ok {
http.Error(w, "max_output_tokens is not supported", http.StatusBadRequest)
return
}
if reqBody["stream"] != true {
http.Error(w, "stream must be true", http.StatusBadRequest)
return
+18 -11
View File
@@ -25,13 +25,14 @@ const (
)
type providerSelection struct {
providerType providerType
apiKey string
apiBase string
proxy string
model string
workspace string
connectMode string
providerType providerType
apiKey string
apiBase string
proxy string
model string
workspace string
connectMode string
enableWebSearch bool
}
func createClaudeAuthProvider(apiBase string) (LLMProvider, error) {
@@ -48,7 +49,7 @@ func createClaudeAuthProvider(apiBase string) (LLMProvider, error) {
return NewClaudeProviderWithTokenSourceAndBaseURL(cred.AccessToken, createClaudeTokenSource(), apiBase), nil
}
func createCodexAuthProvider() (LLMProvider, error) {
func createCodexAuthProvider(enableWebSearch bool) (LLMProvider, error) {
cred, err := getCredential("openai")
if err != nil {
return nil, fmt.Errorf("loading auth credentials: %w", err)
@@ -56,7 +57,9 @@ func createCodexAuthProvider() (LLMProvider, error) {
if cred == nil {
return nil, fmt.Errorf("no credentials for openai. Run: picoclaw auth login --provider openai")
}
return NewCodexProviderWithTokenSource(cred.AccessToken, cred.AccountID, createCodexTokenSource()), nil
p := NewCodexProviderWithTokenSource(cred.AccessToken, cred.AccountID, createCodexTokenSource())
p.enableWebSearch = enableWebSearch
return p, nil
}
func resolveProviderSelection(cfg *config.Config) (providerSelection, error) {
@@ -83,6 +86,7 @@ func resolveProviderSelection(cfg *config.Config) (providerSelection, error) {
}
case "openai", "gpt":
if cfg.Providers.OpenAI.APIKey != "" || cfg.Providers.OpenAI.AuthMethod != "" {
sel.enableWebSearch = cfg.Providers.OpenAI.WebSearch
if cfg.Providers.OpenAI.AuthMethod == "codex-cli" {
sel.providerType = providerTypeCodexCLIToken
return sel, nil
@@ -248,6 +252,7 @@ func resolveProviderSelection(cfg *config.Config) (providerSelection, error) {
}
case (strings.Contains(lowerModel, "gpt") || strings.HasPrefix(model, "openai/")) &&
(cfg.Providers.OpenAI.APIKey != "" || cfg.Providers.OpenAI.AuthMethod != ""):
sel.enableWebSearch = cfg.Providers.OpenAI.WebSearch
if cfg.Providers.OpenAI.AuthMethod == "codex-cli" {
sel.providerType = providerTypeCodexCLIToken
return sel, nil
@@ -338,9 +343,11 @@ func CreateProvider(cfg *config.Config) (LLMProvider, error) {
case providerTypeClaudeAuth:
return createClaudeAuthProvider(sel.apiBase)
case providerTypeCodexAuth:
return createCodexAuthProvider()
return createCodexAuthProvider(sel.enableWebSearch)
case providerTypeCodexCLIToken:
return NewCodexProviderWithTokenSource("", "", CreateCodexCliTokenSource()), nil
c := NewCodexProviderWithTokenSource("", "", CreateCodexCliTokenSource())
c.enableWebSearch = sel.enableWebSearch
return c, nil
case providerTypeClaudeCLI:
return NewClaudeCliProvider(sel.workspace), nil
case providerTypeCodexCLI:
+4 -3
View File
@@ -7,6 +7,7 @@ import (
"time"
"github.com/sipeed/picoclaw/pkg/bus"
"github.com/sipeed/picoclaw/pkg/config"
"github.com/sipeed/picoclaw/pkg/cron"
"github.com/sipeed/picoclaw/pkg/utils"
)
@@ -29,9 +30,9 @@ type CronTool struct {
// NewCronTool creates a new CronTool
// execTimeout: 0 means no timeout, >0 sets the timeout duration
func NewCronTool(cronService *cron.CronService, executor JobExecutor, msgBus *bus.MessageBus, workspace string, restrict bool, execTimeout time.Duration) *CronTool {
execTool := NewExecTool(workspace, restrict)
execTool.SetTimeout(execTimeout) // 0 means no timeout
func NewCronTool(cronService *cron.CronService, executor JobExecutor, msgBus *bus.MessageBus, workspace string, restrict bool, execTimeout time.Duration, config *config.Config) *CronTool {
execTool := NewExecToolWithConfig(workspace, restrict, config)
execTool.SetTimeout(execTimeout)
return &CronTool{
cronService: cronService,
executor: executor,
+76 -9
View File
@@ -4,6 +4,7 @@ import (
"bytes"
"context"
"fmt"
"github.com/sipeed/picoclaw/pkg/config"
"os"
"os/exec"
"path/filepath"
@@ -21,16 +22,82 @@ type ExecTool struct {
restrictToWorkspace bool
}
var defaultDenyPatterns = []*regexp.Regexp{
regexp.MustCompile(`\brm\s+-[rf]{1,2}\b`),
regexp.MustCompile(`\bdel\s+/[fq]\b`),
regexp.MustCompile(`\brmdir\s+/s\b`),
regexp.MustCompile(`\b(format|mkfs|diskpart)\b\s`), // Match disk wiping commands (must be followed by space/args)
regexp.MustCompile(`\bdd\s+if=`),
regexp.MustCompile(`>\s*/dev/sd[a-z]\b`), // Block writes to disk devices (but allow /dev/null)
regexp.MustCompile(`\b(shutdown|reboot|poweroff)\b`),
regexp.MustCompile(`:\(\)\s*\{.*\};\s*:`),
regexp.MustCompile(`\$\([^)]+\)`),
regexp.MustCompile(`\$\{[^}]+\}`),
regexp.MustCompile("`[^`]+`"),
regexp.MustCompile(`\|\s*sh\b`),
regexp.MustCompile(`\|\s*bash\b`),
regexp.MustCompile(`;\s*rm\s+-[rf]`),
regexp.MustCompile(`&&\s*rm\s+-[rf]`),
regexp.MustCompile(`\|\|\s*rm\s+-[rf]`),
regexp.MustCompile(`>\s*/dev/null\s*>&?\s*\d?`),
regexp.MustCompile(`<<\s*EOF`),
regexp.MustCompile(`\$\(\s*cat\s+`),
regexp.MustCompile(`\$\(\s*curl\s+`),
regexp.MustCompile(`\$\(\s*wget\s+`),
regexp.MustCompile(`\$\(\s*which\s+`),
regexp.MustCompile(`\bsudo\b`),
regexp.MustCompile(`\bchmod\s+[0-7]{3,4}\b`),
regexp.MustCompile(`\bchown\b`),
regexp.MustCompile(`\bpkill\b`),
regexp.MustCompile(`\bkillall\b`),
regexp.MustCompile(`\bkill\s+-[9]\b`),
regexp.MustCompile(`\bcurl\b.*\|\s*(sh|bash)`),
regexp.MustCompile(`\bwget\b.*\|\s*(sh|bash)`),
regexp.MustCompile(`\bnpm\s+install\s+-g\b`),
regexp.MustCompile(`\bpip\s+install\s+--user\b`),
regexp.MustCompile(`\bapt\s+(install|remove|purge)\b`),
regexp.MustCompile(`\byum\s+(install|remove)\b`),
regexp.MustCompile(`\bdnf\s+(install|remove)\b`),
regexp.MustCompile(`\bdocker\s+run\b`),
regexp.MustCompile(`\bdocker\s+exec\b`),
regexp.MustCompile(`\bgit\s+push\b`),
regexp.MustCompile(`\bgit\s+force\b`),
regexp.MustCompile(`\bssh\b.*@`),
regexp.MustCompile(`\beval\b`),
regexp.MustCompile(`\bsource\s+.*\.sh\b`),
}
func NewExecTool(workingDir string, restrict bool) *ExecTool {
denyPatterns := []*regexp.Regexp{
regexp.MustCompile(`\brm\s+-[rf]{1,2}\b`),
regexp.MustCompile(`\bdel\s+/[fq]\b`),
regexp.MustCompile(`\brmdir\s+/s\b`),
regexp.MustCompile(`\b(format|mkfs|diskpart)\b\s`), // Match disk wiping commands (must be followed by space/args)
regexp.MustCompile(`\bdd\s+if=`),
regexp.MustCompile(`>\s*/dev/sd[a-z]\b`), // Block writes to disk devices (but allow /dev/null)
regexp.MustCompile(`\b(shutdown|reboot|poweroff)\b`),
regexp.MustCompile(`:\(\)\s*\{.*\};\s*:`),
return NewExecToolWithConfig(workingDir, restrict, nil)
}
func NewExecToolWithConfig(workingDir string, restrict bool, config *config.Config) *ExecTool {
denyPatterns := make([]*regexp.Regexp, 0)
enableDenyPatterns := true
if config != nil {
execConfig := config.Tools.Exec
enableDenyPatterns = execConfig.EnableDenyPatterns
if enableDenyPatterns {
if len(execConfig.CustomDenyPatterns) > 0 {
fmt.Printf("Using custom deny patterns: %v\n", execConfig.CustomDenyPatterns)
for _, pattern := range execConfig.CustomDenyPatterns {
re, err := regexp.Compile(pattern)
if err != nil {
fmt.Printf("Invalid custom deny pattern %q: %v\n", pattern, err)
continue
}
denyPatterns = append(denyPatterns, re)
}
} else {
denyPatterns = append(denyPatterns, defaultDenyPatterns...)
}
} else {
// If deny patterns are disabled, we won't add any patterns, allowing all commands.
fmt.Println("Warning: deny patterns are disabled. All commands will be allowed.")
}
} else {
denyPatterns = append(denyPatterns, defaultDenyPatterns...)
}
return &ExecTool{