test Supermemory with LiteLLM Gemini models

servers/unraid/supermemory-poc/README.md

@@ -10,10 +10,12 @@ 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.
-- Local multilingual `Xenova/bge-m3` embeddings at 1024 dimensions.
+- Remote OpenAI-compatible embeddings through HK LiteLLM using
-- One embedding worker and ingest concurrency 1. This avoids treating known
+  `google/gemini-embedding-2` at 1024 dimensions.
-  concurrent-local-embedding instability as a retrieval-quality result.
+- Extraction, summarization, and other intelligent steps use
-- Extraction/summarization uses an operator-supplied OpenAI-compatible LLM.
+  `gemini-3.5-flash-lite` through the same LiteLLM endpoint.
+- 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
   in memory, so RSS must be watched as the sample grows.
 
@@ -24,8 +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`, and optional model/base URL
+2. Add `SUPERMEMORY_IPV4_ADDRESS`, `OPENAI_API_KEY`,
-   overrides to this Project's Arcane environment. Do not put secrets in Git,
+   `OPENAI_BASE_URL=https://api.opfo.xyz/v1`, and
+   `OPENAI_MODEL=gemini-3.5-flash-lite` to this Project's Arcane environment.
+   `OPENAI_API_KEY` is the HK LiteLLM bearer key. Do not put it in Git,
    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.
@@ -45,7 +49,7 @@ image build is required.
 
 The server creates its client bearer token at:
 
-`/mnt/user/appdata/supermemory-poc/data/api-key`
+`/mnt/user/appdata/supermemory-poc/data-google-gemini-embedding-2-1024/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
@@ -77,17 +81,16 @@ Run `poc.py smoke` before importing any sampled production material. It checks
 the v3 document path plus the v4 search, profile, and conversation endpoints
 used by Hermes.
 
-### Current local-embedding blocker
+### Embedding plan and smoke gate
 
-On the unRaid POC host, `server-v0.0.8` remains healthy but its local embedding
+The earlier local `Xenova/bge-m3` experiment is preserved under the old `data`
-worker returns `unknown certificate verification error` when first loading
+directory because embedding plans cannot be changed in place. This deployment
-`Xenova/bge-m3`. Supplying a Mozilla CA bundle through `SSL_CERT_FILE`,
+uses the fresh `data-google-gemini-embedding-2-1024` directory. The one-shot
-`SSL_CERT_DIR`, `NODE_EXTRA_CA_CERTS`, and Debian's standard CA path did not
+`supermemory-smoke` service verifies the locked embedding plan, performs a
-change the error; TLS verification was never disabled. Do not import sampled
+Chinese hybrid-search round trip, checks provenance metadata, and exercises
-Hindsight or document data until either a remote OpenAI-compatible embedding
+the profile and conversation endpoints. It writes a sentinel after success so
-endpoint or a separately approved local embedding sidecar passes the smoke
+ordinary redeploys do not add duplicate smoke documents. Do not import sampled
-test. The server also returns HTTP 409 when deleting documents stuck in the
+Hindsight or document material until this service exits successfully.
-`embedding` state, so failed smoke rows may require a fresh POC data directory.
 
 ## Evaluation guardrails
 

servers/unraid/supermemory-poc/compose.yaml

@@ -53,20 +53,20 @@ services:
       PORT: "6767"
       SUPERMEMORY_DATA_DIR: /var/lib/supermemory
       SUPERMEMORY_DISABLE_TELEMETRY: "1"
-      SUPERMEMORY_EMBEDDING_PROVIDER: local
+      SUPERMEMORY_EMBEDDING_PROVIDER: openai
-      SUPERMEMORY_EMBEDDING_MODEL: Xenova/bge-m3
+      SUPERMEMORY_EMBEDDING_MODEL: google/gemini-embedding-2
       SUPERMEMORY_EMBEDDING_DIMENSIONS: "1024"
-      SUPERMEMORY_LOCAL_EMBEDDING_POOL_SIZE: "1"
+      SUPERMEMORY_EMBEDDING_BASE_URL: ${OPENAI_BASE_URL:?set OPENAI_BASE_URL in Arcane environment}
-      SUPERMEMORY_LOCAL_EMBEDDING_WASM_THREADS: "1"
-      SUPERMEMORY_LOCAL_EMBEDDING_BATCH_SIZE: "4"
       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}
-      OPENAI_BASE_URL: ${OPENAI_BASE_URL:-}
+      OPENAI_BASE_URL: ${OPENAI_BASE_URL:?set OPENAI_BASE_URL in Arcane environment}
-      OPENAI_MODEL: ${OPENAI_MODEL:-gpt-5.1}
+      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 +86,7 @@ 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:/var/lib/supermemory
+      - /mnt/user/appdata/supermemory-poc/data-google-gemini-embedding-2-1024:/var/lib/supermemory
     networks:
       br0:
         ipv4_address: ${SUPERMEMORY_IPV4_ADDRESS:?set an unused LAN address in Arcane environment}
@@ -100,6 +100,34 @@ services:
         max-size: 10m
         max-file: "3"
 
+  supermemory-smoke:
+    image: curlimages/curl:8.16.0@sha256:463eaf6072688fe96ac64fa623fe73e1dbe25d8ad6c34404a669ad3ce1f104b6
+    container_name: supermemory-poc-smoke
+    user: "0:0"
+    restart: "no"
+    depends_on:
+      supermemory:
+        condition: service_healthy
+    network_mode: service:supermemory
+    environment:
+      SUPERMEMORY_URL: http://127.0.0.1:6767
+      SUPERMEMORY_CONTAINER_TAG: hermes_supermemory_lab
+    entrypoint:
+      - /bin/sh
+      - /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
+    security_opt:
+      - no-new-privileges:true
+    cap_drop:
+      - ALL
+    logging:
+      driver: json-file
+      options:
+        max-size: 1m
+        max-file: "2"
+
 networks:
   fetch:
     internal: false

servers/unraid/supermemory-poc/smoke.sh

@@ -0,0 +1,62 @@
+#!/bin/sh
+set -eu
+
+data_dir=/var/lib/supermemory
+sentinel="$data_dir/.smoke-google-gemini-embedding-2-1024.ok"
+base_url="${SUPERMEMORY_URL:-http://127.0.0.1:6767}"
+container_tag="${SUPERMEMORY_CONTAINER_TAG:-hermes_supermemory_lab}"
+
+if [ -f "$sentinel" ]; then
+  printf '%s\n' '{"smoke":"already-passed"}'
+  exit 0
+fi
+
+api_key=$(cat "$data_dir/api-key")
+auth_header="Authorization: Bearer $api_key"
+marker="supermemory-poc-google-embedding-2-$(date +%s)"
+
+plan="$data_dir/embedding-plan.json"
+for _ in $(seq 1 30); do
+  [ -f "$plan" ] && break
+  sleep 1
+done
+grep -q 'openai' "$plan"
+grep -q 'google/gemini-embedding-2' "$plan"
+grep -q '1024' "$plan"
+
+curl --fail --silent --show-error --max-time 60 \
+  -H "$auth_header" \
+  -H 'Content-Type: application/json' \
+  --data "{\"content\":\"$marker：中文跨文档检索测试；海棠计划的下一步是核对原始报告。\",\"containerTag\":\"$container_tag\",\"customId\":\"$marker\",\"metadata\":{\"source\":\"poc-smoke\",\"status\":\"temporary\"}}" \
+  "$base_url/v3/documents" >/tmp/add.json
+
+found=false
+for _ in $(seq 1 90); do
+  curl --fail --silent --show-error --max-time 60 \
+    -H "$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.json
+  if grep -q '核对原始报告' /tmp/search.json && grep -q 'poc-smoke' /tmp/search.json; then
+    found=true
+    break
+  fi
+  sleep 2
+done
+[ "$found" = true ]
+
+curl --fail --silent --show-error --max-time 60 \
+  -H "$auth_header" \
+  -H 'Content-Type: application/json' \
+  --data "{\"containerTag\":\"$container_tag\",\"q\":\"海棠计划\"}" \
+  "$base_url/v4/profile" >/tmp/profile.json
+grep -q '"profile"' /tmp/profile.json
+
+curl --fail --silent --show-error --max-time 60 \
+  -H "$auth_header" \
+  -H 'Content-Type: application/json' \
+  --data "{\"conversationId\":\"$marker\",\"containerTags\":[\"$container_tag\"],\"messages\":[{\"role\":\"user\",\"content\":\"这只是隔离测试。\"},{\"role\":\"assistant\",\"content\":\"不会作为正式事实。\"}],\"metadata\":{\"source\":\"poc-smoke\",\"status\":\"temporary\"}}" \
+  "$base_url/v4/conversations" >/tmp/conversation.json
+
+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"}'