#8675: fix: Gemini batch embeddings state path, enum values, and download URL

## Summary Fixes #5774 - Three bugs preventing Gemini batch embeddings from working for memory indexing: - **State path**: Code read `status.state` but API returns `status.metadata.state` - **Enum values**: Code checked `"SUCCEEDED"` but API returns `"BATCH_STATE_SUCCEEDED"` - **Download URL**: Code used `:download` suffix but API requires `?alt=media` ## Changes - Added `state?: string` to `GeminiBatchStatus.metadata` type - Added `normalizeGeminiBatchState()` helper that: - Checks `metadata.state` first, falls back to `status.state` - Strips `BATCH_STATE_` and `JOB_STATE_` prefixes - Maintains backwards compatibility with unprefixed values - Fixed download URL from `${file}:download` to `${file}?alt=media` ## Tests Completed - [x] `pnpm build` passes - [x] `pnpm check` passes (lint/format) - [x] `pnpm test src/memory/embeddings.test.ts` passes - [x] Manual test with `OPENCLAW_DEBUG_MEMORY_EMBEDDINGS=1 pnpm openclaw memory index --force --verbose` completes successfully 🤖 Generated with [Claude Code](https://claude.ai/code) <!-- greptile_comment --> <h2>Greptile Overview</h2> <h3>Greptile Summary</h3> This PR fixes Gemini async batch embeddings integration by (1) reading the batch state from `status.metadata.state` (with fallback to `status.state`), (2) normalizing prefixed enum values like `BATCH_STATE_SUCCEEDED`/`JOB_STATE_RUNNING` to the unprefixed values the code expects, and (3) adjusting file download URLs from `:download` to `?alt=media`. These changes improve reliability of memory indexing via Gemini batch embeddings by aligning the client code with the API’s actual response schema and download semantics. <h3>Confidence Score: 4/5</h3> - This PR is largely safe to merge and aligns the Gemini batch client with the API’s actual response shapes. - Changes are localized to the Gemini batch embedding flow and primarily adjust parsing/normalization and URL construction. The only noteworthy concern is the empty-string fallback for `outputFileId` in the early-completion branch, which can obscure missing-output issues and is inconsistent with the wait path. - src/memory/batch-gemini.ts (early-completion outputFileId fallback) <!-- greptile_other_comments_section --> _{(2/5) Greptile learns from your feedback when you react with thumbs up/down!} **Context used:** - Context from `dashboard` - CLAUDE.md ([source](https://app.greptile.com/review/custom-context?memory=fd949e91-5c3a-4ab5-90a1-cbe184fd6ce8)) - Context from `dashboard` - AGENTS.md ([source](https://app.greptile.com/review/custom-context?memory=0d0c8278-ef8e-4d6c-ab21-f5527e322f13)) <!-- /greptile_comment -->