feat(extraction): local query DSL for extract_data (#986)

[shaun0927]

Summary

Implements the smallest safe slice of #986 as a stacked PR on top of #1067 (feat/974-schema-aware-extract):

• adds an OpenChrome-local query parser for extract_data
• supports flat object queries and one root list block with scalar child fields
• compiles query input into the existing ExtractionSchema path
• routes query-derived schemas through feat(extraction): schema-aware deterministic extract_data planning (#974) #1067's schema-aware buildExtractionPlan / fieldPlans
• preserves existing schema-based extract_data behavior
• rejects nested non-root query shapes instead of accepting semantics the extractor cannot satisfy

This intentionally does not add AgentQL, API keys, outbound vendor calls, or full AgentQL language compatibility.

Scope / Repo-Fit Review

Before implementation I checked open PRs and found #1067 already owns schema-aware deterministic extraction planning. This PR is therefore stacked on #1067 rather than targeting develop directly, which avoids duplicating or bypassing that work.

The implemented surface remains aligned with OpenChrome's direction:

• local-first MCP tool behavior
• no external LLM/vendor dependency inside the server
• bounded, deterministic parser
• query text is converted to schema/field plans, not interpolated into browser scripts
• richer standard mode and nested query semantics are left to follow-up issues (feat(core): fast/standard extraction modes with benchmark evidence #989+)

Implementation

• src/extraction/query-parser.ts

  • tokenizes/parses the local query subset
  • supports type hints: string, number, integer, boolean, url, date
  • supports field descriptions
  • builds an ExtractionQueryPlan
  • rejects unsupported nested query shapes

• src/tools/extract-data.ts

  • accepts query as an alternative to schema
  • accepts mode: "fast" placeholder only
  • rejects schema + query together
  • infers multiple for { products[] { ... } }
  • records query-specific domain-memory keys
  • keeps schema mode backward compatible

Verification

• npx jest tests/extraction/query-parser.test.ts tests/tools/extract-data.test.ts tests/extraction/plan.test.ts tests/extraction/strategies.test.ts --runInBand
• npm run build
• npm run lint:changed
• Code review lane: APPROVE
• Architecture lane: CLEAR

Real OpenChrome Verification

Not run in this PR because the branch is stacked on #1067 and should be real-verified after #1067 lands or in CI against the stacked branch. The required real verification from #986 remains:

1. Start real OpenChrome MCP.
2. Navigate to a local product/listing fixture.
3. Call extract_data with:

{
  "tabId": "<fixture tab>",
  "query": "{ products[] { product_name product_price(number) product_url(url) } }",
  "mode": "fast"
}

4. Verify valid JSON, numeric price coercion, URL-like output, no external API key, and output smaller than a full read_page mode="dom" depth=8 response.

Stacking / Merge Notes

Base branch: feat/974-schema-aware-extract (#1067)

Merge order should be:

1. feat(extraction): schema-aware deterministic extract_data planning (#974) #1067
2. this PR

After #1067 merges into develop, this branch can be rebased onto updated develop with the same single commit.

Closes #986 after #1067 is merged and this stacked PR lands.

[gemini-code-assist]

Warning

You have reached your daily quota limit. Please wait up to 24 hours and I will start processing your requests again!

[qodo-code-review]

ⓘ You've reached your Qodo monthly free-tier limit. Reviews pause until next month — upgrade your plan to continue now, or link your paid account if you already have one.