build_context returns digest-less observation permalinks that diverge from the search index

[phernandez]

Found during review of the #909 fix.

`src/basic_memory/services/context_service.py:249` builds observation permalinks inline:

```python
generate_permalink(f"{primary_item.permalink}/observations/{obs.category}/{obs.content}")
```

with no 200-char truncation and (after #909 lands) no content digest, so `build_context` returns permalinks that don't match the search index for observations longer than 200 chars. The divergence pre-dates #909 (truncation was added in #446 without updating this call site).

Fix: `obs` is an `Observation` model here — use `obs.permalink` instead of rebuilding the string inline, so there is exactly one place that defines the synthetic observation permalink format.

Also (release-notes task, same area): existing `search_index` rows for >200-char observations keep their old digest-less permalinks until the entity is re-synced or `bm reindex --search` runs. The release notes for the version shipping #909 should mention the reindex.

Refs #909

[claude]

Claude finished @phernandez's task in 3m 52s —— View job

---

Analysis & Triage

Type: Bug | Priority: Medium | Component: MCP/Core | Complexity: Simple

• Read context_service.py to understand the current code
• Apply fix: use obs.permalink instead of inline rebuild
• Add regression test to test_build_context_with_observations
• Add bug label to issue
• Push changes to remote (blocked: GitHub Actions bot lacks push permission to this repo)

---

Root Cause

context_service.py:249 was building observation permalinks inline:

permalink=generate_permalink(
    f"{primary_item.permalink}/observations/{obs.category}/{obs.content}"
),

This bypassed both the 200-char truncation (added in #446) and the upcoming content digest from #909, causing build_context to return permalinks that don't match the search index for observations longer than 200 chars.

Fix

Replace with obs.permalink — the canonical property on the Observation model that already handles truncation (and will handle the digest once #909 lands):

permalink=obs.permalink,

Files changed:

• src/basic_memory/services/context_service.py — one-line fix
• tests/services/test_context_service.py — regression test test_build_context_observation_permalinks_match_model that verifies returned observation permalinks equal obs.permalink from the model

Note on entity eager-loading: The ObservationRepository.get_load_options() already specifies selectinload(Observation.entity), so obs.entity.permalink (accessed via obs.permalink) is already eagerly loaded in find_by_entities — no N+1 risk.

Also noted (release notes)

As mentioned in the issue, the release notes for the version shipping #909 should mention that existing search_index rows for >200-char observations keep their old digest-less permalinks until the entity is re-synced or bm reindex --search runs.

---

Push blocked: The GitHub Actions bot doesn't have push access to this repository. The commit is ready locally. A maintainer can cherry-pick or apply the patch manually.