separate Supermemory RAG stability gate

This commit is contained in:
Codex
2026-09-09 02:35:21 -07:00
parent 239998de66
commit 6d1dcb6a1a
5 changed files with 211 additions and 25 deletions
+14 -6
View File
@@ -91,9 +91,17 @@ diagnostic exposed the original key in its startup log. The one-shot `supermemor
service verifies the locked embedding plan, tracks the exact document ID to a
terminal state, observes legacy v3 search, gates on v4 document and Chinese
hybrid search, checks provenance metadata, and exercises the profile and
conversation endpoints. It writes a sentinel after success so ordinary
redeploys do not add duplicate smoke documents. Do not import sampled
Hindsight or document material until this service exits successfully.
conversation endpoints. A `/v4/conversations` 404 is reported as a Hermes
compatibility failure but does not invalidate the document-RAG checks. It
writes a sentinel after the document path succeeds so ordinary redeploys do
not add duplicate smoke documents.
The dependent `supermemory-stability` service uses five isolated synthetic
documents instead of Hindsight data. It exercises Chinese, English,
mixed-language, numeric, and preference-like content; checks processing,
exact document retrieval, hybrid retrieval, and provenance metadata; and
reports conversation compatibility separately. The suite passes only when
all five document cases pass and then writes its own sentinel.
On 2026-09-08, the previous remote plan was accepted and locked as
`openai / google/gemini-embedding-2 / 1024`; the LiteLLM embedding and chat
@@ -108,9 +116,9 @@ The v4 `documents` and `hybrid` searches returned the expected Chinese marker
and provenance metadata, while legacy `/v3/search` returned no results. The
self-hosted server-v0.0.8 binary returned 404 for `/v4/conversations`, even
though the self-hosting overview describes that route as part of the full API.
Keep conversation ingest as a failing smoke gate: document RAG can proceed in
isolation, but Hermes auto-capture/session ingest is not validated on this
binary.
Treat conversation ingest as a separate Hermes compatibility gate: document
RAG can proceed in isolation, but Hermes auto-capture/session ingest is not
validated on this binary.
## Evaluation guardrails