Bright Data Local Search
Run powerful web searches locally using Bright Data's SERP API. This skill sets up the unfancy-search pipeline — a local search engine with query expansion, multi-engine SERP retrieval, RRF reranking, deduplication, and domain clustering.
Important: This skill uses the LOCAL version only. Do not use the hosted endpoint.
Prerequisites
- Bright Data account with SERP API access
- Anthropic API key (for query expansion, optional)
- Docker (recommended) or Node.js 18+ with pnpm
Setup
Step 1: Clone and Configure
git clone https://github.com/yaronbeen/unfancy-search.git
cd unfancy-search
cp .env.example .envStep 2: Set Environment Variables
Edit .env with your credentials:
BRIGHT_DATA_API_TOKEN=your_brightdata_token
BRIGHT_DATA_SERP_ZONE=serp_api1
ANTHROPIC_API_KEY=your_anthropic_key # Optional: enables AI query expansionGet your Bright Data token from: https://brightdata.com (SERP API section)
Step 3: Start the Local Server
Docker (recommended):
docker compose up -d
# Server runs at http://localhost:3000Node.js:
pnpm install
pnpm dev
# Server runs at http://localhost:3000API Endpoints
All requests go to http://localhost:3000:
| Endpoint | Method | Description |
|---|---|---|
/api/search |
POST | Start a search job |
/api/search-status/{jobId} |
GET | Poll for results |
/api/baseline |
POST | Trigger baseline collection |
/api/baseline-status/{id} |
GET | Poll baseline progress |
Running a Search
Step 1: Submit Search
curl -X POST http://localhost:3000/api/search \
-H "Content-Type: application/json" \
-d '{"query": "your search term"}'The response returns a jobId.
Step 2: Poll for Results
curl http://localhost:3000/api/search-status/{jobId}Poll every 3 seconds until status is "done".
Search Parameters
| Parameter | Type | Default | Description |
|---|---|---|---|
query |
string | required | Search query |
expand |
boolean | false |
Enable AI query expansion via Claude |
research |
boolean | false |
Research mode (12 sub-queries for max coverage) |
engines |
string[] | all | SERP engines to use |
geo |
string | — | Geographic region filter |
count |
number | 10 | Max results (up to 10) |
includeDomains |
string[] | — | Only include results from these domains |
excludeDomains |
string[] | — | Exclude results from these domains |
Search Modes
- Basic (
expand: false): Single query, fastest, no AI cost - Expanded (
expand: true): Claude Haiku generates 3 sub-queries for broader coverage - Research (
research: true): 12 sub-queries for maximum coverage
Usage Examples
Basic Search from an Agent
# Start search
JOB_ID=$(curl -s -X POST http://localhost:3000/api/search \
-H "Content-Type: application/json" \
-d '{"query": "best practices for API rate limiting"}' | jq -r '.jobId')
# Poll until done
while true; do
RESULT=$(curl -s http://localhost:3000/api/search-status/$JOB_ID)
STATUS=$(echo $RESULT | jq -r '.status')
if [ "$STATUS" = "done" ]; then
echo $RESULT | jq '.results'
break
fi
sleep 3
doneResearch Mode with Domain Filtering
curl -X POST http://localhost:3000/api/search \
-H "Content-Type: application/json" \
-d '{
"query": "kubernetes scaling strategies",
"research": true,
"excludeDomains": ["pinterest.com", "quora.com"]
}'Adding Search to an Existing Agent
To give your Claude Code agent search capabilities:
- Ensure the local server is running (
docker compose up -din the unfancy-search directory) - Your agent can use
curlorfetchto queryhttp://localhost:3000/api/search - Parse the ranked results to ground responses with real web data
Response Format
Results include:
- Ranked URLs with RRF scores
- Domain clustering (grouped by source)
- Cost transparency (per-search expense breakdown)
- Raw and unique result counts
- Search duration
Troubleshooting
| Issue | Solution |
|---|---|
| Server won't start | Verify Docker is running or Node.js 18+ installed |
| No results returned | Check BRIGHT_DATA_API_TOKEN is valid and SERP API zone is active |
| Query expansion not working | Verify ANTHROPIC_API_KEY is set and valid |
| Slow responses | Disable expand mode for faster single-query searches |
| Port 3000 in use | Stop other services or modify the port in docker-compose.yml |
Scanned by
NVIDIA SkillSpector
Automated static analysis (no LLM) · v2.12.0 · 2026-11-19
Partial analysis: some checks could not run on this component.
4 observations in 2 categories
- Data Exfiltration 3
- Privilege Escalation 1
SkillSpector looks for patterns such as shell commands, credential access, network calls or unpinned dependencies. An observation is something worth reading in the files before you install, not a confirmed vulnerability. Components that run code locally, like hooks, MCP servers and security tooling, naturally show more of them. Open one to see the lines it points at.
Credential AccessPrivilege Escalation · SKILL.md:25
| 23 | git clone https://github.com/yaronbeen/unfancy-search.git |
| 24 | cd unfancy-search |
| 25 | cp .env.example .env |
| 26 | ``` |
| 27 |
External Transmission ×3Data Exfiltration · SKILL.md:71
| 69 | |
| 70 | ```bash |
| 71 | curl -X POST http://localhost:3000/api/search \ |
| 72 | -H "Content-Type: application/json" \ |
| 73 | -d '{"query": "your search term"}' |
Learn more about NVIDIA SkillSpector or how components are reviewed.
Agent compatibility
Checked by NVIDIA SkillEvaluator Harness Check · v1.0.1 · 2026-11-18
How each agent reads this skill once it is installed, from the agents' documented loading rules. No agent was run. A note means the agent loads the skill but changes part of its text, or drops it.
Claude Code
loadsCodex
loadsCopilot CLI
loadsOpenCode
loadsHermes
loadsOpenClaw
loads
Every agent loads this skill as written. No notes.
Learn more about NVIDIA SkillEvaluator.