feat(server): attachment embedding (#13348)

<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit

* **New Features**
* Added support for managing "blobs" in Copilot context, including
adding and removing blobs via new GraphQL mutations and UI fields.
* Introduced tracking and querying of blob embeddings within workspaces,
enabling search and similarity matching for blob content.
* Extended Copilot context and workspace APIs, schema, and UI to display
and manage blobs alongside existing documents and files.

* **Bug Fixes**
* Updated context and embedding status logic to handle blobs, ensuring
accurate status reporting and embedding management.

* **Tests**
* Added and updated test cases and snapshots to cover blob embedding
insertion, matching, and removal scenarios.

* **Documentation**
* Updated GraphQL schema and TypeScript types to reflect new
blob-related fields and mutations.

* **Chores**
* Refactored and cleaned up code to support new blob entity and
embedding logic, including renaming and updating internal methods and
types.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
This commit is contained in:
DarkSky
2025-07-31 06:07:28 +08:00
committed by GitHub
parent b6a5bc052e
commit feb42e34be
24 changed files with 689 additions and 84 deletions

View File

@@ -147,6 +147,28 @@ export class CopilotContextService implements OnApplicationBootstrap {
return null;
}
async matchWorkspaceBlobs(
workspaceId: string,
content: string,
topK: number = 5,
signal?: AbortSignal,
threshold: number = 0.5
) {
if (!this.embeddingClient) return [];
const embedding = await this.embeddingClient.getEmbedding(content, signal);
if (!embedding) return [];
const blobChunks = await this.models.copilotWorkspace.matchBlobEmbedding(
workspaceId,
embedding,
topK * 2,
threshold
);
if (!blobChunks.length) return [];
return await this.embeddingClient.reRank(content, blobChunks, topK, signal);
}
async matchWorkspaceFiles(
workspaceId: string,
content: string,