convoslisted

# Convos Retrieve with these commands: ```bash convos query "natural language question" -n 8 -f jsonl # conceptual/paraphrased discovery; default when wording is uncertain convos search "exact terms" -n 8 -c 160 -f jsonl # known terms, quotes, ids, filenames; one hit/conversation convos read abc123 -n 20 -c 2000 -f jsonl # bounded recent context from one known conversation convos read abc123 --around MESSAGE_ID -n 20 -f jsonl # bounded context around an exact discovery hit convos related MESSAGE_ID -n 5 -c 160 -f jsonl # semantic neighbors of one exact turn convos trail MESSAGE_ID --depth 2 --width 3 -f json # bounded multi-hop evidence graph convos redact scan -f json # local high-confidence secret locations, never values convos redact status -f json # audit automatic pre-encryption team redactions convos resume /path/to/project -f json # live Git plus bounded exact project handoff convos replay CONVERSATION_ID -f json # exact bounded messages, tools, and edits convos sql "SELECT ..." -f jsonl # structured filters, joins, counts, and history ``` Output: add `-f jsonl` (stream, one JSON object per line) or `-f json` (array); default is human text. Prefer `jsonl` when parsing programmatically. Schema (write `convos sql` against these tables): - `conversations(id, source, title, create