feat(artifact): add ResetKnowledgeBaseEmbeddingsAdmin RPC for BM25 migration (#722)

**Because**

- Existing default knowledge bases in Zilliz Cloud may not have BM25
native hybrid search support enabled in their Milvus collection schema
- The `convert000070` migration needs a way to reset KB embeddings so
that new Milvus collections with BM25-enabled schema can be created
- There was no admin API to trigger a full embedding reset (drop
collection + clear PostgreSQL data + reset file statuses)

**This commit**

- Adds `ResetKnowledgeBaseEmbeddingsAdmin` RPC to
`ArtifactPrivateService` that resets all embeddings for a knowledge base
- Adds `ResetKnowledgeBaseEmbeddingsAdminRequest` and
`ResetKnowledgeBaseEmbeddingsAdminResponse` message types in
`knowledge_base.proto`
- Updates comment in `file.proto` to reflect hash-based ID format
(`col-xxx`) instead of UUID
- Applies formatting improvements (line breaks) to long RPC declarations
for better readability

Author: pinglin

artifact/v1alpha/artifact_private_service.proto

@@ -39,7 +39,7 @@ service ArtifactPrivateService {
   //
   // Updates a file allowing system-reserved tag prefixes like "agent:".
   // Used by agent-backend to set collection association tags (e.g.,
-  // "agent:collection:{uid}").
+  // "agent:collection:{id}" where {id} is hash-based like col-xxx).
   rpc UpdateFileAdmin(UpdateFileAdminRequest) returns (UpdateFileAdminResponse);
 
   // Get Object (admin only)
@@ -122,4 +122,7 @@ service ArtifactPrivateService {
 
   // Get the current default system configuration (admin only)
   rpc GetDefaultSystemAdmin(GetDefaultSystemAdminRequest) returns (GetDefaultSystemAdminResponse);
+
+  // Reset knowledge base embeddings (admin only)
+  rpc ResetKnowledgeBaseEmbeddingsAdmin(ResetKnowledgeBaseEmbeddingsAdminRequest) returns (ResetKnowledgeBaseEmbeddingsAdminResponse);
 }

artifact/v1alpha/file.proto

@@ -542,7 +542,7 @@ message ReprocessFileResponse {
 // UpdateFileAdminRequest represents a request to update a file with
 // system-reserved tags (admin only). Used by internal services like
 // agent-backend to set tags with reserved prefixes (e.g.,
-// "agent:collection:{uid}").
+// "agent:collection:{id}" where {id} is hash-based like col-xxx).
 // Follows AIP-134: https://google.aip.dev/134
 message UpdateFileAdminRequest {
   // The file resource to update. The file's `name` field identifies the

artifact/v1alpha/knowledge_base.proto

@@ -314,3 +314,23 @@ message UpdateKnowledgeBaseAdminResponse {
   // The updated knowledge base resource.
   KnowledgeBase knowledge_base = 1 [(google.api.field_behavior) = OUTPUT_ONLY];
 }
+
+// ResetKnowledgeBaseEmbeddingsAdminRequest represents a request to reset all
+// embeddings for a knowledge base (admin only).
+message ResetKnowledgeBaseEmbeddingsAdminRequest {
+  // The resource name of the knowledge base to reset embeddings for.
+  // Format: `namespaces/{namespace}/knowledge-bases/{knowledge_base}`
+  string name = 1 [
+    (google.api.field_behavior) = REQUIRED,
+    (google.api.resource_reference) = {type: "api.instill.tech/KnowledgeBase"}
+  ];
+}
+
+// ResetKnowledgeBaseEmbeddingsAdminResponse represents a response for resetting
+// knowledge base embeddings.
+message ResetKnowledgeBaseEmbeddingsAdminResponse {
+  // The knowledge base that had its embeddings reset.
+  KnowledgeBase knowledge_base = 1 [(google.api.field_behavior) = OUTPUT_ONLY];
+  // Number of files that will be re-embedded.
+  int32 files_to_reembed = 2 [(google.api.field_behavior) = OUTPUT_ONLY];
+}