Quickstart¶

This guide assumes you have installed the package and started SearXNG.

CLI commands¶

The llm-context CLI has three commands, each exposing a different stage of the pipeline.

`search` - URL discovery¶

Searches via SearXNG and prints results. No page fetching.

llm-context search "Python asyncio best practices"

llm-context search "Python asyncio" --max-results 20 --json

`collect` - fetch and extract¶

Fetches pages, extracts main content, shows per-source status.

llm-context collect "Python asyncio best practices" --max-sources 5

llm-context collect "Python asyncio" --verbose   # show URLs + details

`build` - full pipeline¶

Runs the complete pipeline and returns context ready for an LLM prompt.

llm-context build "Python asyncio best practices" --budget 4000

# Save to file
llm-context build "Python asyncio" -o context.md

# Output JSON (for scripting)
llm-context build "Python asyncio" --json | jq .token_estimate

# Verbose: show source table + context
llm-context build "Python asyncio" --verbose --budget 6000

Common options¶

All commands share these flags:

Flag	Default	Description
`--searxng-url` / `-u`	`http://localhost:8888`	SearXNG base URL
`--language` / `-l`	`en`	Search language (`pl`, `de`, `fr`, …)
`--max-results`	`10`	Results fetched from SearXNG
`--max-sources`	`5`	Pages to fetch and extract
`--max-passages`	`12`	Passages included in output
`--budget`	`4000`	Token budget for context
`--timeout`	`10.0`	HTTP timeout in seconds
`--no-cache`	off	Disable in-memory TTL cache
`--json`	off	Output raw JSON
`--output` / `-o`	-	Write to file
`--verbose` / `-v`	off	Show detailed source info

Environment variables¶

export SEARXNG_URL=http://localhost:8888
llm-context build "my query"   # no need to pass --searxng-url

Example: feed context to an LLM¶

# Build context and pipe into any LLM CLI
CONTEXT=$(llm-context build "how does Python GIL work" --budget 3000)
echo "$CONTEXT" | llm "Summarise the above in 3 bullet points"

Or save to a file and reference it:

llm-context build "how does Python GIL work" -o /tmp/context.md
llm -f /tmp/context.md "What are the main points?"