Intent vs. source: fix the config or fix the doc? #63

Closed
opened 2026-09-24 19:12:22 +00:00 by haylan · 1 comment
Owner

Part of #55

Question

The audit (#56) found docs stating an intent the source violates. Under 'source is truth' the doc would just be rewritten to match — unless the source is the bug. For each, decide: change the source, or document the source as-is?

  • OmniRoute dashboard port: docs say never published; compose publishes it.
  • qdrant (6333, no auth) and neo4j ports published to the LAN; .env.example claims 'not exposed off-box'.
  • Lazytainer idle-stop is presented as working, but it never fires (#40) — keep lazytainer, or remove it?

HITL: /grilling.

Part of #55 ## Question The audit (#56) found docs stating an *intent* the source violates. Under 'source is truth' the doc would just be rewritten to match — unless the source is the bug. For each, decide: change the source, or document the source as-is? - OmniRoute dashboard port: docs say never published; compose publishes it. - qdrant (6333, no auth) and neo4j ports published to the LAN; .env.example claims 'not exposed off-box'. - Lazytainer idle-stop is presented as working, but it never fires (#40) — keep lazytainer, or remove it? HITL: /grilling.
haylan added the wayfinder:grilling label 2026-09-24 19:12:22 +00:00
haylan added a new dependency 2026-09-24 19:12:23 +00:00
haylan self-assigned this 2026-09-24 19:22:39 +00:00
Author
Owner

Resolution

  1. OmniRoute dashboard port — source is right, docs are wrong. It stays published on the LAN; it is simply not routed to the internet (no reverse-proxy host, no port-forward). Docs change: drop every 'never published' claim and the container-IP / docker inspect workaround in the onboarding doc; reach the dashboard at http://<server>:<dashboard port> from the LAN.
  2. qdrant / neo4j — source changes:
    • neo4j is removed from the stack (service, neo4j-data volume, NEO4J_PASSWORD in .env.example + update.sh, network-access doc section).
    • qdrant stays, bound to loopback only (127.0.0.1:6333:6333) — no API key needed. Fix the .env.example 'not exposed off-box' claim to match.
  3. Lazytainer is removed (service, the lazytainer.* labels on llama-server, LAZYTAINER_INACTIVE_TIMEOUT, README 'auto-suspending' line). Its idle-stop could never fire (see 'Why lazytainer idle-stop doesn't work today' #40). Docs describe the main backend as always running.

Manual server-side steps for the deploy list: delete the neo4j-data volume and the knowledge.proxy-ai.home reverse-proxy host; stale NEO4J_PASSWORD / LAZYTAINER_* lines in the server's .env are harmless.

## Resolution 1. **OmniRoute dashboard port** — source is right, docs are wrong. It stays published on the LAN; it is simply not routed to the internet (no reverse-proxy host, no port-forward). Docs change: drop every 'never published' claim and the container-IP / `docker inspect` workaround in the onboarding doc; reach the dashboard at `http://<server>:<dashboard port>` from the LAN. 2. **qdrant / neo4j** — source changes: - **neo4j is removed** from the stack (service, `neo4j-data` volume, `NEO4J_PASSWORD` in .env.example + update.sh, network-access doc section). - **qdrant stays, bound to loopback only** (`127.0.0.1:6333:6333`) — no API key needed. Fix the .env.example 'not exposed off-box' claim to match. 3. **Lazytainer is removed** (service, the `lazytainer.*` labels on llama-server, `LAZYTAINER_INACTIVE_TIMEOUT`, README 'auto-suspending' line). Its idle-stop could never fire (see 'Why lazytainer idle-stop doesn't work today' #40). Docs describe the main backend as always running. Manual server-side steps for the deploy list: delete the `neo4j-data` volume and the `knowledge.proxy-ai.home` reverse-proxy host; stale `NEO4J_PASSWORD` / `LAZYTAINER_*` lines in the server's .env are harmless.
Sign in to join this conversation.
1 Participants
Notifications
Due Date
No due date set.
Blocks
#60 Target doc structure
haylan/LLM-Server
Reference: haylan/LLM-Server#63