Compaction and pages

The server never writes prose. Compaction is a queue of work items that an agent, or a person, completes. The server checks the result before storing it.

Work items

compact, through the tool or memors-mcp compact, scans a namespace and creates items:

Kind	Created when
`page`	An entity has 3 or more live passages and no page
`stale`	A page’s sources changed or were forgotten, or the entity gained 2 or more passages since the page was built
`conflict`	Two live facts about one subject overlap in time and disagree
`merge`	An open merge candidate exists
`duplicate`	Two passages from different documents are near duplicates (word 3-gram Jaccard of at least 0.75)

Each item’s payload carries everything needed to do it: the passages, the facts and the previous page. No second call is needed.

Submitting

memors-mcp compact --ns default --kinds page,stale --json
memors-mcp submit 42 --content-file page.md --title "Quorum replication" --dry-run
memors-mcp submit 42 --content-file page.md --title "Quorum replication"
memors-mcp submit 43 --keep memo://fact/0199…
memors-mcp submit 44 --accept            # merge item
memors-mcp submit 45 --skip --reason "not worth a page"

Every page submission reports:

diff: a line diff against the previous page.
omission check: recorded facts about the subject whose content words are mostly absent from the page. Less than 75% coverage is flagged.
corruption check: page sentences that share fewer than half of their content words with any source.

Pages are stored as is_inference = 1, with the passages they cite. They are searchable with granularity: page, and marked stale when a source changes. Raw rows are never modified.

Lint

memors-mcp lint --json

Reports contradictions, orphan entities, missing pages, stale pages and expired facts, with addresses. Lint changes nothing.

The optional Ollama executor

For unattended page writing from the terminal:

MEMORS_OLLAMA_MODEL=qwen2.5:7b-instruct memors-mcp compact --kinds page --executor ollama          # dry run
MEMORS_OLLAMA_MODEL=qwen2.5:7b-instruct memors-mcp compact --kinds page --executor ollama --apply  # submit

The endpoint is MEMORS_OLLAMA_URL, default http://127.0.0.1:11434. It must be loopback unless --allow-remote is passed.
Use a non-thinking instruction model. Every page still goes through the same checks.
No MCP code path can reach the executor.

Observing it

memors_kb_work_items_open, memors_kb_pages, memors_kb_pages_stale, memors_store_work_items_total{kind,event} and memors_store_pages_marked_stale_total.