fix(honcho): warn about minimal reasoning_level's 250-token output cap

[LavyaTandel]

What

Expanded the reasoning_level parameter description in REASONING_SCHEMA to document minimal's hard 250-token output cap and steer models to low+ for anything beyond single-fact lookups.

Why

Honcho's minimal dialectic tier hard-caps output at 250 tokens — intended for single fast factual lookups. On reasoning-capable models, hidden reasoning tokens share that budget, so synthesis queries truncate before producing a final answer. The schema gave no signal this would happen.

Fix

Schema-string-only change in plugins/memory/honcho/__init__.py:

• Tool description: added '250-token cap — single facts only' to minimal
• Parameter description: added warning about truncation behavior

No behavioral code modified.

Runtime Proof

Before: model chose minimal for multi-fact query → truncated output
After: same query with updated schema → model chose low unprompted → complete answer

Duplicate Scan

No existing PRs for #59470.

Closes #59470

[alt-glitch]

This was generated by AI during triage.

Duplicate of #59472 (earlier, same fix). Both edit the reasoning_level parameter description in REASONING_SCHEMA (plugins/memory/honcho/__init__.py) to warn that minimal has a 250-token cap and steer to low+ — schema-string-only, same code site. #59472 was opened first, so it's canonical.

[teknium1]

Thanks for the narrowly scoped schema update. Current main still exposes reasoning_level as a model-selected parameter and forwards it unchanged through plugins/memory/honcho/__init__.py:1352-1362; the added guidance therefore reaches the relevant model decision point.

The member triage comment already identifies open PR #59472 as the earlier canonical implementation of the same mitigation. Its patch overlaps this PR at the same parameter description, so maintainers should reconcile the wording rather than apply both independently.

Automated hermes-sweeper review.

[teknium1]

Superseded on main by #66052 (commit b170f52), which merged the earliest submission for #59470 — #59472 by @vizi0uz, submitted ~40 min before this one. The parameter-guide text on main now carries the 250-token minimal-cap warning your PR also added. Your top-level description tweak was a nice touch we may fold into a future polish pass. Thanks @LavyaTandel, sorry it collided with an earlier duplicate.