fire_rag 1.3.0

Firebase RAG for AI Agents

1.3.0

• Added repository-oriented ingestion and retrieval APIs with Repository, RepositoryRecord, FireResult, and FireResultSet.
• Added FireRag.querySingle(...) and FireRag.query(...) for nearest-neighbor search scoped to a repository.
• Added generation-aware refresh utilities with pushRecordVectors(...), deleteRecordVectors(...), and stale-result filtering during query.
• Added ChunkSource variants for Cloud Storage files, Firestore documents, and inline strings.
• Added record progress tracking for core and distilled vector work through coreVectorProgress and distilledVectorProgress.
• Added image-aware embedding support in TaskEmbed for image-modality chunks stored in the configured bucket.
• Updated FireRag.init(...) to require vectorCollection and bucket.
• Updated the README to document the new repository, refresh, and query workflow.

1.1.2

• ID Fixes

1.1.1

• Generated models

1.1.0

• Updated chunk persistence to use the serialized agentic.Chunk document shape instead of hand-built field maps.
• Stored record, up, down, and vector on the top-level chunk document, while keeping app-specific values inside metadata.
• Updated TaskEmbed to rebuild text from Chunk.fullContent before embedding.
• Updated TaskDistill to read and write Firestore records through the Chunk model path and preserve chunk metadata during recursive distillation.
• Updated package documentation to describe the current Firestore schema and how it fits with rag.

1.0.0

• Added FireRag.init() bootstrap wiring for a shared embedding model, chat model, and TaskManager.
• Added Cloud Task executors for chunking, recursive distillation, and embedding work.
• Added TaskChunk to download source files, create persisted base chunks, and fan out embedding batches.
• Added TaskDistill to build higher-LOD distilled chunks, maintain parent/child linkage fields, and continue recursive summarization until a single output remains.
• Added TaskEmbed to read stored chunk text and write vector embeddings back to Firestore documents.
• Added CloudRecursiveChunkingTaskManager as the package entry point for recursive ingestion workflows.