Commit 3999944
fix(ollama): worker reaches ollama via service name + bump timeouts 30s→90s
Two distinct prod failures behind the 15% vocab-enrichment success rate:
1. Worker was hitting http://localhost:11434 because docker-compose only
set Ollama__BaseUrl on the api service, not the worker. Without the
env override, appsettings.json's "http://localhost:11434" stayed in
effect — every BookMetadataGenerator call logged "Connection refused".
Fix: add Ollama__BaseUrl + Ollama__Model to the worker env block,
mirroring the api block.
2. API was reaching Ollama fine (http://ollama:11434) but timing out at
30s — gemma4:e4b is CPU-only on this box (no GPU), 100% CPU, first
inference after model load is comfortably north of 30s. Fix: bump
Ollama:TimeoutSeconds 30 → 90 in both api and worker appsettings.
Net effect on the article: distractor/hint/explanation success rate
should climb from 15% to ~100% once deployed. Will re-collect stats
after rollout before publishing.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>1 parent 26186b7 commit 3999944
3 files changed
Lines changed: 8 additions & 2 deletions
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
45 | 45 | | |
46 | 46 | | |
47 | 47 | | |
48 | | - | |
| 48 | + | |
49 | 49 | | |
50 | 50 | | |
51 | 51 | | |
| |||
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
12 | 12 | | |
13 | 13 | | |
14 | 14 | | |
15 | | - | |
| 15 | + | |
16 | 16 | | |
17 | 17 | | |
18 | 18 | | |
| |||
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
110 | 110 | | |
111 | 111 | | |
112 | 112 | | |
| 113 | + | |
| 114 | + | |
| 115 | + | |
| 116 | + | |
| 117 | + | |
| 118 | + | |
113 | 119 | | |
114 | 120 | | |
115 | 121 | | |
| |||
0 commit comments