CLI Reference

Every command supports --json for machine-readable output on stdout. Human-readable output goes to stderr.

Commands

Command	Description
lore	Launch interactive TUI
lore init	Initialize .lore/ repository
lore ingest <path|url>	Ingest file or URL into raw/
lore compile [--force]	Compile changed raw sources into wiki articles (hash-based incremental, lock-guarded)
lore index [--repair]	Rebuild FTS5 index + index.md (optional manifest repair from raw/)
lore query "<q>" [--no-file-back] [--normalize-question]	BFS/DFS + LLM Q&A
lore search "<term>" [--limit N]	FTS5/BM25 search
lore path "<A>" "<B>"	Shortest path between articles
lore explain "<concept>"	Deep-dive on a concept
lore lint	Wiki health checks + structured diagnostics
lore watch	Watch raw/ and wiki/ and auto-compile raw changes
lore angela [install|run]	Git commit capture
lore export <format> [--out dir]	Export wiki
lore mcp	Start MCP server
lore status	Repo health dashboard
lore settings	Configure API keys and model

Common Flags and Behaviors

• --json: structured machine output on stdout.
• human mode: operational summaries on stderr; primary text output on stdout where relevant.
• lore ingest --json: includes duplicate indicator when content already exists.
• lore compile: uses hash-based incremental compile, skipping unchanged extracted content via manifest.json extractedHash values.
• lore compile --force: bypasses hash skipping and recompiles all valid raw entries.
• lore compile: guarded by .lore/compile.lock to prevent concurrent runs.
• lore index --repair: reconstructs missing manifest entries before rebuild.
• lore lint --json: includes line-aware diagnostics[] entries with rule, severity, file, optional line, and message.
• lore watch: debounces raw changes, runs compile automatically, and queues one additional pass when changes land during compile.
• lore query --normalize-question: conservative typo cleanup while preserving technical tokens.

Guide links:

• lore init: Quickstart
• lore ingest: Ingesting Content
• lore compile: Compiling Your Wiki
• lore search: Searching and Querying
• lore query: Searching and Querying
• lore explain: Explain Command
• lore lint: Linting and Health
• lore watch: Compiling Your Wiki
• lore export: Exporting
• lore angela: Lore Angela
• lore mcp: MCP Server
• lore settings: Configuration
• troubleshooting workflows: Troubleshooting
• repository operating patterns: Best Practices

Examples:

lore ingest ./docs/architecture.md --json
lore index --repair --json
lore query "teh qurey about src/core/mcp.ts" --normalize-question --json

Settings Command

Interactive mode:

lore settings

Non-interactive mode:

lore settings list [--scope global|repo|all] [--json]
lore settings get [key] [--scope global|repo|all] [--json]
lore settings set <key> <value> [--scope global|repo|all]
lore settings unset <key> [--scope global|repo|all]

Common keys:

• Global: openrouterApiKey, replicateApiToken, cloudflareAccountId, cloudflareToken
• Repo: model, temperature, maxTokens (optional), webExporter

Notes:

• lore settings unset maxTokens --scope repo removes the explicit token cap.
• When maxTokens is unset, Lore omits max_tokens in LLM requests.

Run Logs

• lore ingest, lore compile, and lore query create JSONL logs in .lore/logs/<runId>.jsonl.
• Human mode prints run start/end summaries (including runId and log path) to stderr.
• JSON mode includes runId and logPath in command output.
• Logs are rotated automatically; configure retention with LORE_LOG_MAX_FILES.

Notes for Automation

• Prefer --json for scripting.
• Prefer lore lint --json and treat diagnostics severity=error findings as hard failures.
• For deterministic maintenance pipelines, run in this order:
  • lore ingest ...
  • lore compile
  • lore index --repair
  • lore lint --json

Exit Codes

• 0 -- success
• 1 -- error