test Supermemory with native Gemini

servers/unraid/supermemory-poc/.env.example

@@ -1,8 +1,6 @@
 # Set only in the unRaid Arcane Project environment. Never commit the real file.
 SUPERMEMORY_IPV4_ADDRESS=
 
-# Extraction/summarization LLM. OPENAI_BASE_URL may point to an approved
+# Native Gemini powers extraction/summarization and remote embeddings.
-# OpenAI-compatible endpoint; keep it blank for the OpenAI API.
+# Keep the real key only in this Project's Arcane environment.
-OPENAI_API_KEY=
+GEMINI_API_KEY=
-OPENAI_BASE_URL=
-OPENAI_MODEL=gpt-5.1

servers/unraid/supermemory-poc/README.md

@@ -10,10 +10,10 @@ production Hindsight service or change the default Hermes profile.
   cache live under `/mnt/user/appdata/supermemory-poc/data`.
 - A dedicated address on Docker `br0`; port `6767` is not published on the
   unRaid host address.
-- Remote OpenAI-compatible embeddings through HK LiteLLM using
+- Native Gemini embeddings using `gemini-embedding-2` at 768 dimensions.
-  `google/gemini-embedding-2` at 1024 dimensions.
+- Extraction, summarization, and other intelligent steps use Supermemory's
-- Extraction, summarization, and other intelligent steps use
+  native Gemini provider default. This cleanly removes the LiteLLM and
-  `gemini-3.5-flash-lite` through the same LiteLLM endpoint.
+  OpenAI-compatible translation layer from the diagnostic path.
 - Ingest concurrency remains 1 so the first POC emphasizes correctness and
   observable cost over throughput.
 - Resource ceiling: 4 CPUs and 8 GiB RAM. Supermemory Local keeps its corpus
@@ -26,11 +26,10 @@ directory and re-ingest when comparing a different embedding plan.
 
 1. Confirm the chosen `SUPERMEMORY_IPV4_ADDRESS` is absent from both Arcane's
    `br0` attachments and the LAN neighbor/DHCP tables.
-2. Add `SUPERMEMORY_IPV4_ADDRESS`, `OPENAI_API_KEY`,
+2. Add `SUPERMEMORY_IPV4_ADDRESS` and `GEMINI_API_KEY` to this Project's
-   `OPENAI_BASE_URL=https://api.opfo.xyz/v1`, and
+   Arcane environment. Do not put the key in Git, Compose, activity notes, or
-   `OPENAI_MODEL=gemini-3.5-flash-lite` to this Project's Arcane environment.
+   chat. Legacy `OPENAI_*` values may remain stored in Arcane for rollback,
-   `OPENAI_API_KEY` is the HK LiteLLM bearer key. Do not put it in Git,
+   but this Compose file does not inject them into the container.
-   Compose, activity notes, or chat.
 3. Confirm `/mnt/user/appdata` has room for the 298 MiB server binary, the
    multilingual model cache, data, and rollback copy.
 4. Confirm no other build, image pull, backup, or migration is active on the
@@ -49,7 +48,7 @@ image build is required.
 
 The server creates its client bearer token at:
 
-`/mnt/user/appdata/supermemory-poc/data-google-gemini-embedding-2-1024/api-key`
+`/mnt/user/appdata/supermemory-poc/data-gemini-embedding-2-768/api-key`
 
 Read it through an authorized unRaid/Arcane console without printing it into
 logs. Save it only as `SUPERMEMORY_API_KEY` in the `supermemory-lab` Hermes
@@ -83,24 +82,23 @@ used by Hermes.
 
 ### Embedding plan and smoke gate
 
-The earlier local `Xenova/bge-m3` experiment is preserved under the old `data`
+The earlier local `Xenova/bge-m3` and LiteLLM-backed embedding experiments are
-directory because embedding plans cannot be changed in place. This deployment
+preserved under their old data directories because embedding plans cannot be
-uses the fresh `data-google-gemini-embedding-2-1024` directory. The one-shot
+changed in place. This deployment uses the fresh
-`supermemory-smoke` service verifies the locked embedding plan, performs a
+`data-gemini-embedding-2-768` directory. The one-shot `supermemory-smoke`
-Chinese hybrid-search round trip, checks provenance metadata, and exercises
+service verifies the locked embedding plan, tracks the exact document ID to a
-the profile and conversation endpoints. It writes a sentinel after success so
+terminal state, separately checks v3 search, v4 document search, and Chinese
-ordinary redeploys do not add duplicate smoke documents. Do not import sampled
+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.
 
-On 2026-09-08, the remote plan was accepted and locked as
+On 2026-09-08, the previous remote plan was accepted and locked as
-`openai / google/gemini-embedding-2 / 1024`; the LiteLLM embedding and
+`openai / google/gemini-embedding-2 / 1024`; the LiteLLM embedding and chat
-`gemini-3.5-flash-lite` chat endpoints also passed independently. Supermemory
+endpoints also passed independently. Supermemory listed a completed smoke
-listed the smoke document and reported a completed document in the test
+document, but both v4 and v3 searches returned no hit. The old smoke mixed
-container, but exact searches for its unique marker returned no hit through
+list-wide status checks and did not track the newly created document, so this
-either `/v4/search` or `/v3/search` (`supermemory-smoke` exit 44). Treat this
+native Gemini run uses a new store and a stricter document-specific gate.
-as a self-hosted indexing/retrieval blocker. Do not import sampled material or
-enable the Hermes lab profile until a newer server build or upstream fix makes
-the same smoke pass.
 
 ## Evaluation guardrails
 

servers/unraid/supermemory-poc/compose.yaml

@@ -53,20 +53,15 @@ services:
       PORT: "6767"
       SUPERMEMORY_DATA_DIR: /var/lib/supermemory
       SUPERMEMORY_DISABLE_TELEMETRY: "1"
-      SUPERMEMORY_EMBEDDING_PROVIDER: openai
+      SUPERMEMORY_EMBEDDING_PROVIDER: gemini
-      SUPERMEMORY_EMBEDDING_MODEL: google/gemini-embedding-2
+      SUPERMEMORY_EMBEDDING_MODEL: gemini-embedding-2
-      SUPERMEMORY_EMBEDDING_DIMENSIONS: "1024"
+      SUPERMEMORY_EMBEDDING_DIMENSIONS: "768"
-      SUPERMEMORY_EMBEDDING_BASE_URL: ${OPENAI_BASE_URL:?set OPENAI_BASE_URL in Arcane environment}
       SUPERMEMORY_EMBEDDING_RAM_LIMIT: 2gb
       SUPERMEMORY_INGEST_CONCURRENCY: "1"
       SSL_CERT_FILE: /opt/supermemory/bin/ca-certificates.crt
       SSL_CERT_DIR: /etc/ssl/certs
       NODE_EXTRA_CA_CERTS: /opt/supermemory/bin/ca-certificates.crt
-      OPENAI_API_KEY: ${OPENAI_API_KEY:?set OPENAI_API_KEY in Arcane environment}
+      GEMINI_API_KEY: ${GEMINI_API_KEY:?set GEMINI_API_KEY in Arcane environment}
-      OPENAI_BASE_URL: ${OPENAI_BASE_URL:?set OPENAI_BASE_URL in Arcane environment}
-      OPENAI_MODEL: ${OPENAI_MODEL:-gemini-3.5-flash-lite}
-      OPENAI_FAST_MODEL: ${OPENAI_MODEL:-gemini-3.5-flash-lite}
-      OPENAI_TEXT_MODEL: ${OPENAI_MODEL:-gemini-3.5-flash-lite}
     entrypoint:
       - /bin/bash
       - -ec
@@ -86,7 +81,7 @@ services:
     volumes:
       - /mnt/user/appdata/supermemory-poc/bin:/opt/supermemory/bin:ro
       - /mnt/user/appdata/supermemory-poc/bin/ca-certificates.crt:/etc/ssl/certs/ca-certificates.crt:ro
-      - /mnt/user/appdata/supermemory-poc/data-google-gemini-embedding-2-1024:/var/lib/supermemory
+      - /mnt/user/appdata/supermemory-poc/data-gemini-embedding-2-768:/var/lib/supermemory
     networks:
       br0:
         ipv4_address: ${SUPERMEMORY_IPV4_ADDRESS:?set an unused LAN address in Arcane environment}
@@ -117,7 +112,7 @@ services:
       - /opt/supermemory-poc/smoke.sh
     volumes:
       - ./smoke.sh:/opt/supermemory-poc/smoke.sh:ro
-      - /mnt/user/appdata/supermemory-poc/data-google-gemini-embedding-2-1024:/var/lib/supermemory
+      - /mnt/user/appdata/supermemory-poc/data-gemini-embedding-2-768:/var/lib/supermemory
     security_opt:
       - no-new-privileges:true
     cap_drop:

servers/unraid/supermemory-poc/smoke.sh

@@ -2,7 +2,7 @@
 set -eu
 
 data_dir=/var/lib/supermemory
-sentinel="$data_dir/.smoke-google-gemini-embedding-2-1024.ok"
+sentinel="$data_dir/.smoke-gemini-embedding-2-768.ok"
 base_url="${SUPERMEMORY_URL:-http://127.0.0.1:6767}"
 container_tag="${SUPERMEMORY_CONTAINER_TAG:-hermes_supermemory_lab}"
 
@@ -13,7 +13,7 @@ fi
 
 api_key=$(cat "$data_dir/api-key")
 auth_header="Authorization: Bearer $api_key"
-marker="supermemory-poc-google-embedding-2-$(date +%s)"
+marker="supermemory-poc-gemini-native-$(date +%s)"
 
 fail() {
   code="$1"
@@ -27,35 +27,10 @@ for _ in $(seq 1 30); do
   [ -f "$plan" ] && break
   sleep 1
 done
-grep -q 'openai' "$plan" || fail 11 embedding-plan-provider
+grep -q 'gemini' "$plan" || fail 11 embedding-plan-provider
-grep -q 'google/gemini-embedding-2' "$plan" || fail 12 embedding-plan-model
+grep -q 'gemini-embedding-2' "$plan" || fail 12 embedding-plan-model
-grep -q '1024' "$plan" || fail 13 embedding-plan-dimensions
+grep -q '768' "$plan" || fail 13 embedding-plan-dimensions
 
-if ! curl --fail --silent --show-error --max-time 60 \
-  -H "$auth_header" \
-  -H 'Content-Type: application/json' \
-  --data "{\"containerTags\":[\"$container_tag\"],\"limit\":20}" \
-  "$base_url/v3/documents/list" >/tmp/documents.json; then
-  fail 21 documents-list
-fi
-
-if grep -q 'supermemory-poc-google-embedding-2' /tmp/documents.json; then
-  if grep -Eq '"status"[[:space:]]*:[[:space:]]*"failed"' /tmp/documents.json; then
-    fail 31 document-processing-failed
-  fi
-  if ! grep -Eq '"status"[[:space:]]*:[[:space:]]*"done"' /tmp/documents.json; then
-    if grep -Eq '"status"[[:space:]]*:[[:space:]]*"embedding"' /tmp/documents.json; then
-      fail 32 document-stuck-embedding
-    fi
-    if grep -Eq '"status"[[:space:]]*:[[:space:]]*"chunking"' /tmp/documents.json; then
-      fail 33 document-stuck-chunking
-    fi
-    if grep -Eq '"status"[[:space:]]*:[[:space:]]*"indexing"' /tmp/documents.json; then
-      fail 34 document-stuck-indexing
-    fi
-    fail 35 document-not-done
-  fi
-else
 if ! curl --fail --silent --show-error --max-time 60 \
   -H "$auth_header" \
   -H 'Content-Type: application/json' \
@@ -63,42 +38,57 @@ else
   "$base_url/v3/documents" >/tmp/add.json; then
   fail 22 document-add
 fi
-fi
 
-existing_marker=$(grep -o 'supermemory-poc-google-embedding-2-[0-9]*' /tmp/documents.json | head -n 1)
+document_id=$(sed -n 's/.*"id"[[:space:]]*:[[:space:]]*"\([^"]*\)".*/\1/p' /tmp/add.json | head -n 1)
-[ -n "$existing_marker" ] || fail 36 document-marker-missing
+[ -n "$document_id" ] || fail 23 document-id-missing
+
+document_status=
+for _ in $(seq 1 120); do
+  if ! curl --fail --silent --show-error --max-time 60 \
+    -H "$auth_header" \
+    "$base_url/v3/documents/$document_id" >/tmp/document.json; then
+    fail 24 document-get
+  fi
+  if grep -Eq '"status"[[:space:]]*:[[:space:]]*"failed"' /tmp/document.json; then
+    fail 31 document-processing-failed
+  fi
+  if grep -Eq '"status"[[:space:]]*:[[:space:]]*"done"' /tmp/document.json; then
+    document_status=done
+    break
+  fi
+  sleep 2
+done
+[ "$document_status" = done ] || fail 35 document-not-done
+
+if ! curl --fail --silent --show-error --max-time 15 \
+  -H "$auth_header" \
+  -H 'Content-Type: application/json' \
+  --data "{\"q\":\"$marker\",\"containerTag\":\"$container_tag\"}" \
+  "$base_url/v3/search" >/tmp/search-v3.json; then
+  fail 41 search-v3-request
+fi
+grep -q "$marker" /tmp/search-v3.json || fail 44 search-v3-no-hit
 
 curl --fail --silent --show-error --max-time 15 \
   -H "$auth_header" \
   -H 'Content-Type: application/json' \
-  --data "{\"q\":\"$existing_marker\",\"containerTag\":\"$container_tag\",\"searchMode\":\"hybrid\",\"limit\":5}" \
+  --data "{\"q\":\"$marker\",\"containerTag\":\"$container_tag\",\"searchMode\":\"documents\",\"limit\":5}" \
-  "$base_url/v4/search" >/tmp/search-exact.json || fail 41 search-exact-request
+  "$base_url/v4/search" >/tmp/search-documents.json || fail 42 search-documents-request
-if ! grep -q "$existing_marker" /tmp/search-exact.json; then
+grep -q "$marker" /tmp/search-documents.json || fail 45 search-documents-no-hit
-  curl --fail --silent --show-error --max-time 15 \
+grep -q 'poc-smoke' /tmp/search-documents.json || fail 46 search-documents-metadata-missing
-    -H "$auth_header" \
-    -H 'Content-Type: application/json' \
-    --data "{\"q\":\"$existing_marker\",\"containerTag\":\"$container_tag\"}" \
-    "$base_url/v3/search" >/tmp/search-exact-v3.json || fail 43 search-exact-v3-request
-  if grep -q "$existing_marker" /tmp/search-exact-v3.json; then
-    grep -q 'poc-smoke' /tmp/search-exact-v3.json || fail 49 search-v3-metadata-missing
-    fail 48 search-v3-only
-  fi
-  fail 44 search-exact-no-hit
-fi
-grep -q 'poc-smoke' /tmp/search-exact.json || fail 45 search-exact-metadata-missing
 
 curl --fail --silent --show-error --max-time 15 \
   -H "$auth_header" \
   -H 'Content-Type: application/json' \
   --data "{\"q\":\"海棠计划下一步是什么？\",\"containerTag\":\"$container_tag\",\"searchMode\":\"hybrid\",\"limit\":5}" \
-  "$base_url/v4/search" >/tmp/search-semantic.json || fail 42 search-semantic-request
+  "$base_url/v4/search" >/tmp/search-hybrid.json || fail 47 search-hybrid-request
-grep -q '核对原始报告' /tmp/search-semantic.json || fail 46 search-semantic-no-hit
+grep -q '核对原始报告' /tmp/search-hybrid.json || fail 48 search-hybrid-no-hit
-grep -q 'poc-smoke' /tmp/search-semantic.json || fail 47 search-semantic-metadata-missing
+grep -q 'poc-smoke' /tmp/search-hybrid.json || fail 49 search-hybrid-metadata-missing
 
 curl --fail --silent --show-error --max-time 60 \
   -H "$auth_header" \
   -H 'Content-Type: application/json' \
-  --data "{\"containerTag\":\"$container_tag\",\"q\":\"海棠计划\"}" \
+  --data "{\"containerTag\":\"$container_tag\",\"q\":\"$marker\"}" \
   "$base_url/v4/profile" >/tmp/profile.json || fail 51 profile-request
 grep -q '"profile"' /tmp/profile.json || fail 52 profile-shape
 
@@ -109,4 +99,4 @@ curl --fail --silent --show-error --max-time 60 \
   "$base_url/v4/conversations" >/tmp/conversation.json || fail 61 conversation-ingest
 
 touch "$sentinel"
-printf '%s\n' '{"embedding_plan":"openai/google/gemini-embedding-2/1024","llm":"gemini-3.5-flash-lite","chinese_hybrid_search":"ok","metadata_traceability":"ok","profile":"ok","conversation_ingest":"accepted"}'
+printf '%s\n' '{"embedding_plan":"gemini/gemini-embedding-2/768","llm":"gemini-native-default","v3_search":"ok","v4_documents_search":"ok","chinese_hybrid_search":"ok","metadata_traceability":"ok","profile":"ok","conversation_ingest":"accepted"}'