feat: Add SearXNG as web search provider

[truongvinht]

📝 Description

Adds SearXNG as a third web search backend alongside DuckDuckGo and Brave.

Motivation:

• DuckDuckGo blocks datacenter IPs (Oracle Cloud, AWS, etc.), returning empty/blocked responses (138 bytes)
• Brave Search eliminated their free API tier (Feb 12, 2026), now requiring $5/month minimum
• SearXNG is self-hosted, free, and aggregates 70+ search engines with a native JSON API

Configurable via config.json. Fully backwards compatible — existing configs without the searxng block work unchanged.

🗣️ Type of Change

• 🐞 Bug fix (non-breaking change which fixes an issue)
• ✨ New feature (non-breaking change which adds functionality)
• 📖 Documentation update
• ⚡ Code refactoring (no functional changes, no api changes)

🤖 AI Code Generation

• 🤖 Fully AI-generated (100% AI, 0% Human)
• 🛠️ Mostly AI-generated (AI draft, Human verified/modified)
• 👨‍💻 Mostly Human-written (Human lead, AI assisted or none)

🔗 Related Issue

No existing issue. DuckDuckGo IP blocking and Brave free tier removal affect all users deploying PicoClaw on cloud VMs.

📚 Technical Context (Skip for Docs)

• Reference URL: https://searxng.org / https://docs.searxng.org/dev/search_api.html
• Reasoning: SearXNG provides a self-hosted meta-search engine with JSON API support (/search?q=query&format=json), making it the only free and reliable web search option for datacenter-hosted PicoClaw instances.

🧪 Test Environment

• Hardware: Oracle Cloud E2.1.Micro (1 OCPU, 1GB RAM)
• OS: Ubuntu 22.04
• Model/Provider: DeepSeek V3
• Channels: Discord

📸 Evidence (Optional)

Click to view Logs/Screenshots

Feb 20 13:56:05 instance-20260215-1129 picoclaw[2394]: 2026/02/20 13:56:05 [2026-02-20T13:56:05Z] [INFO] agent: LLM requested tool calls {tools=[web_search], count=1, iteration=1, agent_id=main}
Feb 20 13:56:05 instance-20260215-1129 picoclaw[2394]: 2026/02/20 13:56:05 [2026-02-20T13:56:05Z] [INFO] agent: Tool call: web_search({"count":10,"query":"OpenClaw AI agent framework news 2026"}) {iteration=1, agent_id=main, tool=web_search}
Feb 20 13:56:05 instance-20260215-1129 picoclaw[2394]: 2026/02/20 13:56:05 [2026-02-20T13:56:05Z] [INFO] tool: Tool execution started {tool=web_search, args=map[count:10 query:OpenClaw AI agent framework news 2026]}
Feb 20 13:56:06 instance-20260215-1129 picoclaw[2394]: 2026/02/20 13:56:06 [2026-02-20T13:56:06Z] [INFO] tool: Tool execution completed {tool=web_search, duration_ms=1163, result_length=3152}
Feb 20 13:56:20 instance-20260215-1129 picoclaw[2394]: 2026/02/20 13:56:20 [2026-02-20T13:56:20Z] [INFO] agent: Response: Based on the search results, here's the latest news about OpenClaw:....

☑️ Checklist

• My code/docs follow the style of this project.
• I have performed a self-review of my own changes.
• I have updated the documentation accordingly.

[sky5454]

I think adding this will be convenient for users to aggregate their own information, but at the same time, we may also need to consider the hidden danger of config unlimited expansion. Maybe we can extract the tool or tool-web item of the configuration file and make it into a separate configuration file?

[IceWreck]

++ Having this would be great - lots of people using picoclaw would also be self hosting searxng.

[truongvinht]

Update: fix merge conflict

[airpods69]

@truongvinht I did try this out with my locally hosted instance (built your branch)
but getting a 403 error :(

[airpods69]

okay nvm found it
https://stackoverflow.com/a/79056789

The problem is that by default the SearXNG seems to only support format=html. I don't know what is the best way to resolve this but you can get format=json working by editing /etc/searxng/settings.yml and adding json to the enabled search formats. Would be nice to have environment variable for this for the docker image.

[nikolasdehor]

Well-motivated feature. DuckDuckGo blocking datacenter IPs and Brave removing their free tier are real problems for self-hosted deployments. SearXNG fills the gap nicely.

The implementation is clean and follows the existing search provider pattern. A few minor observations:

1. New HTTP client per request: Same pattern as other providers in this file, but http.Client{Timeout: 10 * time.Second} should be created once and reused. Not a blocker since the other providers do the same thing.

2. No proxy support: Unlike Brave and Perplexity providers, SearXNG does not use the opts.Proxy setting. Since SearXNG is typically self-hosted on the same network, this is usually fine, but for consistency it should either support proxy or document why it does not.

3. README changes are extensive: The PR updates the README in multiple places with search provider comparisons and configuration examples. The content is accurate and helpful. The Brave Search description was updated from "Free tier available" to "Paid ($5/1000 queries)" which is consistent with the stated motivation.

4. Priority order: SearXNG is placed after Brave but before Tavily in the priority chain. This makes sense since SearXNG is self-hosted (presumably more reliable than Tavily for the user) but does not require an API key like Brave.

5. Perplexity config added to README: The PR also adds Perplexity configuration examples to the README sections where only Brave/DuckDuckGo/Tavily were listed. This is a good completeness fix but unrelated to SearXNG.

LGTM.

[afjcjsbx]

LGTM

[afjcjsbx]

could you fix lint pls?

[truongvinht]

could you fix lint pls?

I just fixed merge conflict and lint issue

[afjcjsbx]

sorry, again 😞

[CLAassistant]

All committers have signed the CLA.

[truongvinht]

@afjcjsbx dont worry just fixed conflict

[Orgmar]

@truongvinht Cảm ơn bạn đã thêm SearXNG làm backend tìm kiếm web! Phân tích vấn đề rất rõ ràng, DuckDuckGo chặn IP datacenter còn Brave đã bỏ gói miễn phí, SearXNG tự host được là giải pháp rất hợp lý.

Chúng mình đang xây dựng PicoClaw Dev Group trên Discord để các contributors trao đổi và hợp tác. Nếu bạn quan tâm, hãy gửi email đến [email protected] với tiêu đề [Join PicoClaw Dev Group] truongvinht, mình sẽ gửi link mời Discord cho bạn!