fix(context): re-probe context length on session reset (#31043)

[Tranquil-Flow]

What

CLI /new (and /reset) do not refresh context_compressor.context_length after provider config changes. After adjusting LM Studio's context length and reloading the model, /new reuses the stale cached value — the agent operates with the wrong context budget.

Fix

ContextCompressor.on_session_reset() now re-probes context length from the provider. For LM Studio specifically, it invalidates the short-lived endpoint metadata cache before probing, so /new reflects a just-reloaded model instead of a 5-minute-old /api/v1/models response.

• agent/model_metadata.py — +9: invalidate_endpoint_model_metadata_cache(base_url) public helper
• agent/context_compressor.py — +26: on_session_reset() re-probes + LM Studio cache invalidation

Why over #31067

A competing PR (#31067) identifies the same endpoint-cache issue, but:

• Patches cli.py only (not the reset abstraction used by /new and /reset)
• Imports private model_metadata cache internals directly
• Recreates ContextCompressor instead of reusing the existing reset hook
• Has no committed regression tests

This fix is narrower in the runtime reset abstraction and better covered.

Tests

python3 -m pytest tests/agent/test_context_compressor.py -q -o addopts=
87 passed in 25s

4 regression tests: re-probe on reset, no-op when unchanged, probe failure retains stale value, LM Studio cache invalidation.

Closes #31043

[teknium1]

Thanks for targeting the existing compressor lifecycle instead of replacing the context engine. The premise remains valid on current main: CLI /new retains the agent and calls reset_session_state() (cli.py:7073-7076), whose lifecycle calls on_session_reset() (run_agent.py:647-651); the built-in reset currently does not re-resolve context_length (agent/context_compressor.py:726-747).

Problems

• The proposed resolver call omits config_context_length. Current initialization stores the resolved override (agent/agent_init.py:1747-1749), and resolver step 0 makes that value authoritative (agent/model_metadata.py:2051-2053). A reset could overwrite an explicit configured budget.
• The proposed outer exception handler does not preserve the old value for normal probe-down behavior: failed custom/local probes fall through to catalog/default resolution (agent/model_metadata.py:2212-2258, 2404-2426).

Suggested changes

• Rework the reset hook on current main to preserve configured overrides and dynamically refresh only auto-detected values.
• Add coverage for explicit override preservation and provider-unavailable fallback, alongside the changed LM Studio value case.

Automated hermes-sweeper review.

[teknium1]

get_model_context_length() gives config_context_length precedence, but this reset call omits it. Preserve the resolved override used at agent initialization; otherwise /new can replace an explicit model.context_length or custom-provider override with endpoint metadata.