Solution for long term memory for agent coding CLIs and to facilitate handoff between different agent vendors
Rust
docs(importer): note how reruns treat pages over 64,000 characters
João Matheus4193f776db64
diff --git a/companions/ai-memory-importer/README.md b/companions/ai-memory-importer/README.mdindex b943b48a..a8b5adc7 100644--- a/companions/ai-memory-importer/README.md+++ b/companions/ai-memory-importer/README.md@@ -137,7 +137,10 @@ The page records the source file's basename, format, SHA-256 digest, andextractor version, followed by sanitized text. Its path includes the fullsource digest, and its last line carries a deterministic provenance marker.A rerun therefore performs no duplicate write. If another page already owns-that path, the import refuses to replace it.+that path, the import refuses to replace it. The rerun compares the whole+stored page, read through `memory_read_page`, which returns at most 64,000+characters of a body; a page longer than that cannot be compared, so its+rerun fails with that reason instead of being reported as present.This is text extraction, not a media pipeline. Image-only PDFs need OCR beforeimport. Images, raw audio/video, and proprietary media transcription are not@@ -230,7 +233,8 @@ The agentmemory routes keep the same dry-run default, `--apply` plus- Each record has a stable ingest key, a UUIDv5 of its agentmemory id, writtenon the page's last line. A rerun compares the entire sanitized planned body,including that line, with the page the server already holds before skipping- it. Editing the body while keeping its attribution footer therefore cannot+ it. `memory_read_page` returns at most 64,000 characters of a body, so a+ longer page cannot be compared and stops the run with that reason. Editing the body while keeping its attribution footer therefore cannotauthorize a later imported version to replace the edit. Server-sidesanitation that changes the body also stops automatic resumption. Thiscomparison covers the body, not frontmatter-only edits; a later imported