fix: migrate legacy llm_compress_keep_recent config

[iuiu-py]

Fixes #8598

Motivation

Upgrades could leave the legacy provider_settings.llm_compress_keep_recent count in place without populating the replacement ratio. The recent-context preference was therefore lost when the new compression configuration was used.

Changes

• Convert the legacy recent-message count to keep_recent_ratio using the legacy max_context_length denominator.
• Clamp the migrated ratio to the supported 0.0-0.3 range and fall back to 0.15 when the legacy count is invalid or cannot be expressed against an unbounded/invalid limit.
• Preserve an existing ratio when both keys are present. For an already-migrated local runner, populate a missing runner ratio before removing the legacy input.
• Add coverage for legacy-only migration, both-key precedence, invalid legacy values, and preserving an existing runner ratio.

Testing

• uv run pytest tests/unit/test_agent_runner_config.py -q - 45 passed
• uv run pytest tests/unit/test_agent_runner_config.py tests/unit/test_config.py -q - 116 passed
• uv run ruff format --check .
• uv run ruff check .
• git diff --check

Summary by Sourcery

Preserve recent-context compression preferences when migrating legacy agent runner configuration.

Bug Fixes:

• Migrate the legacy recent-message compression count into the new bounded keep-recent ratio so upgrades preserve context preferences.

Enhancements:

• Preserve explicitly configured ratios and populate missing ratios in existing local agent runners while removing legacy settings.
• Use a safe default for invalid or unsupported legacy values and cap migrated ratios at the supported maximum.

Tests:

• Add coverage for legacy count conversion, ratio precedence, invalid-value fallback, and existing runner migration.

[sourcery-ai]

Hey - I've found 1 issue

Prompt for AI Agents

Please address the comments from this code review:

## Individual Comments

### Comment 1
<location path="astrbot/core/utils/migra_helper.py" line_range="164-169" />
<code_context>
+    if legacy_count_value is None:
+        return _DEFAULT_KEEP_RECENT_RATIO
+
+    try:
+        legacy_count = int(legacy_count_value)
+    except (TypeError, ValueError):
+        return _DEFAULT_KEEP_RECENT_RATIO
+    if isinstance(legacy_count_value, bool) or legacy_count < 0:
+        return _DEFAULT_KEEP_RECENT_RATIO
</code_context>
<issue_to_address>
**issue (bug_risk):** A boolean `max_context_length` is converted to `1` instead of being rejected, so `True` causes the legacy count to be divided by one and commonly produces the maximum ratio `0.3` instead of the invalid-limit fallback `0.15`.

**Triggers:** When the legacy configuration contains a boolean `max_context_length`.

**Suggested fix:** Reject boolean values for `max_context_length` before converting it to an integer.

```suggestion
    if isinstance(provider_settings.get("max_context_length", -1), bool):
        return _DEFAULT_KEEP_RECENT_RATIO
    try:
        max_turns = int(provider_settings.get("max_context_length", -1))
    except (TypeError, ValueError):
        return _DEFAULT_KEEP_RECENT_RATIO
    if max_turns <= 0:
        return _DEFAULT_KEEP_RECENT_RATIO
```
</issue_to_address>

Sourcery assessment

Needs a human reviewer. 1 finding to address first, and if the count-to-ratio conversion is wrong, the migrated compression ratio is persisted in the agent-runner configuration and remains after reverting the code, so subsequent compression behavior can be incorrect. The setting is bounded and can be repaired by updating or re-running configuration migration; it does not delete data or trigger an irreversible external action.

Blocking findings: astrbot/core/utils/migra_helper.py:169

---

Sourcery is free for open source - if you like our reviews please consider sharing them ✨

• X
• Mastodon
• LinkedIn
• Facebook

[sourcery-ai]

issue (bug_risk): A boolean max_context_length is converted to 1 instead of being rejected, so True causes the legacy count to be divided by one and commonly produces the maximum ratio 0.3 instead of the invalid-limit fallback 0.15.

Triggers: When the legacy configuration contains a boolean max_context_length.

Suggested fix: Reject boolean values for max_context_length before converting it to an integer.

Suggested change

	try:
	max_turns = int(provider_settings.get("max_context_length", -1))
	except (TypeError, ValueError):
	return _DEFAULT_KEEP_RECENT_RATIO
	if max_turns <= 0:
	return _DEFAULT_KEEP_RECENT_RATIO
	if isinstance(provider_settings.get("max_context_length", -1), bool):
	return _DEFAULT_KEEP_RECENT_RATIO
	try:
	max_turns = int(provider_settings.get("max_context_length", -1))
	except (TypeError, ValueError):
	return _DEFAULT_KEEP_RECENT_RATIO
	if max_turns <= 0:
	return _DEFAULT_KEEP_RECENT_RATIO