feat(search) Add lightweight query planner config for intent analysis

[heaoxiang-ai]

Description

Adds an optional dedicated query planner model configuration for OpenViking retrieval intent analysis. When configured, search intent analysis uses this lightweight planner instead of the general VLM; when omitted or empty, OpenViking preserves the existing behavior and falls back to vlm.

The PR keeps the default example config unchanged to avoid implying that every deployment should run a local intent-analysis model. The opt-in planner setup is documented in the configuration guide instead.

Related Issue

N/A

Type of Change

• Bug fix
• New feature
• Breaking change
• Documentation update
• Refactoring
• Test update

Changes Made

• Added top-level query_planner config support using the existing VLMConfig shape.
• Added OpenVikingConfig.get_query_planner() to centralize fallback to vlm when query_planner is unset or empty.
• Updated IntentAnalyzer to call the configured query planner for retrieval intent analysis and query planning.
• Documented query_planner in English and Chinese configuration guides as an optional opt-in setting.
• Added tests/retrieve/test_intent_analyzer_query_planner.py for config parsing and planner-vs-VLM routing behavior.
• Added scripts/test_ov_intent_planner.py as a local smoke test for validating query planner wiring against representative intent-analysis scenarios.

Testing

• python -m pytest tests/retrieve/test_intent_analyzer_query_planner.py passed: 4 passed, 4 warnings.
• Local smoke test was run against an Ollama-backed query planner and returned typed query plans for operational, informational, and memory-constrained scenarios, plus an empty query list for a conversational scenario.

Checklist

• Code follows the existing project style.
• I have performed a self-review of the changes.
• I have run relevant local verification.
• I have added or updated automated tests where appropriate.
• I have updated documentation where appropriate.

Screenshots

N/A

Additional Notes

The new config is backward-compatible: existing deployments without query_planner continue using vlm for intent analysis.

[github-actions]

PR Reviewer Guide 🔍

(Review updated until commit 1655053)

Here are some key observations to aid the review process:

⏱️ Estimated effort to review: 2 🔵🔵⚪⚪⚪

🏅 Score: 95

🧪 PR contains tests

🔒 No security concerns identified

✅ No TODO sections

🔀 No multiple PR themes

⚡ No major issues detected

[github-actions]

PR Code Suggestions ✨

Explore these optional code suggestions:

Category	Suggestion	Impact
General	Simplify query planner selection Simplify query planner selection by directly using config.get_query_planner(), which already implements the fallback logic to vlm. The conditional check and manual fallback are now redundant. openviking/retrieve/intent_analyzer.py [65-70] config = get_openviking_config() -if hasattr(config, "get_query_planner"): - query_planner = config.get_query_planner() -else: - query_planner = getattr(config, "query_planner", None) or config.vlm +query_planner = config.get_query_planner() response = await query_planner.get_completion_async(prompt) Suggestion importance[1-10]: 6 __ Why: The PR added OpenVikingConfig.get_query_planner() which already implements fallback to vlm, so the conditional check and manual fallback are redundant. Simplifying to directly call config.get_query_planner() makes the code cleaner and more maintainable.	Low

[github-actions]

Persistent review updated to latest commit 1655053

[github-actions]

PR Code Suggestions ✨

No code suggestions found for the PR.