Releases: strawberry-code/ragify
Releases · strawberry-code/ragify
v2.0.0 - Pipeline Optimization
Breaking Changes
- Unica immagine Docker: rimossi
Dockerfile.tikaeDockerfile.tika.local, ora esiste soloDockerfilecon Tika integrato - Tika sempre obbligatorio: rimosso flag
--no-tikae logica opzionale, Tika server sempre attivo - Tag immagine semplificato: usare
ghcr.io/strawberry-code/ragify:latest(rimosso suffisso-tika)
Added
- Dynamic batching: nuovo sistema di batching basato su token budget invece di batch size fisso
- EMBEDDING_TOKEN_BUDGET: nuova env var (default 1800) per controllare token massimi per batch
- Index file_hash: creazione automatica index su Qdrant per query O(1) invece di scroll O(N)
- FileHashCache: cache in-memory per evitare query ripetute durante indicizzazione
- Tika server mode: Tika avviato come server all'avvio container (porta 9998), elimina cold start 5-10s per file
Performance
| Metrica | Prima | Dopo |
|---|---|---|
| Chiamate embedding (1k chunk) | ~334 | ~50-100 |
| Tika cold start | 5-10s/file | 0s |
| Hash check | O(N) scroll | O(1) index |
Docker
docker pull ghcr.io/strawberry-code/ragify:2.0.0
docker pull ghcr.io/strawberry-code/ragify:latestv1.3.2
What's Changed
Fixed
- Further reduced default
EMBEDDING_BATCH_SIZEfrom 10 to 3 - This prevents batch embedding failures that caused slow fallback to single-chunk mode
- Rule:
batch_size × avg_chunk_tokensmust be < 2048 (nomic-embed-text context limit)
Docker Images
ghcr.io/strawberry-code/ragify:1.3.2-tikaghcr.io/strawberry-code/ragify:latest-tika
v1.3.1
What's Changed
Fixed
- Reduced default
EMBEDDING_BATCH_SIZEfrom 32 to 10 to prevent Ollama "cannot decode batches" errors with large uploads - New
EMBEDDING_BATCH_SIZEenv var allows tuning for different Ollama configurations
Docker Images
ghcr.io/strawberry-code/ragify:1.3.1-tikaghcr.io/strawberry-code/ragify:latest-tika
Note: This image doesn't have the embedding model pre-cached. It will be downloaded automatically on first startup (~274MB).
v1.3.0
What's Changed
Changed
- Frontend semplificato: rimosso upload cartelle, ora richiede file ZIP per caricare multiple documenti
- Rimossa dipendenza JSZip dal frontend (compressione ora gestita dall'utente)
Added
- Modale informativa quando si tenta di trascinare una cartella, con istruzioni per creare ZIP
Docker Images
ghcr.io/strawberry-code/ragify:1.3.0-tikaghcr.io/strawberry-code/ragify:latest-tika
v1.2.3
What's Changed
Fixed
- Pinned Ollama to v0.11.0 in Dockerfiles to avoid embedding bugs in 0.12.x/0.13.x (see ollama#13054)
- Added missing
num_ctxoption to batch embedding API call (/api/embed) - Restored
max_tokensin config.yaml to 2048 (nomic-embed-text context limit)
Docker Images
ghcr.io/strawberry-code/ragify:1.2.3-tikaghcr.io/strawberry-code/ragify:latest-tika
v1.2.2
What's Changed
Added
- Chunking parameters now configurable via environment variables:
CHUNK_SIZE(default 400) andCHUNK_MAX_TOKENS(default 1500)
Fixed
- Reduced default chunk size to prevent Ollama panic with nomic-embed-text model (2048 token context limit)
Docker Images
ghcr.io/strawberry-code/ragify:1.2.2-tikaghcr.io/strawberry-code/ragify:latest-tika
v1.2.1
What's Changed
Fixed
- File cleanup now guaranteed via finally block: uploaded files are deleted after Qdrant upload or on pipeline failure
Docker Images
ghcr.io/strawberry-code/ragify:1.2.1-tikaghcr.io/strawberry-code/ragify:latest-tika
v1.2.0
What's Changed
Added
- Progress bar now shows processing stages (extracting, chunking, embedding, uploading) in real-time
- Client-side ZIP compression for multi-file uploads, reducing N HTTP requests to 1 and improving upload speed
- New
/api/upload-zipendpoint for server-side ZIP extraction and batch processing - Browser-side progress feedback: "Zipping..." and "Uploading..." phases before server processing
Changed
- Batch embedding using Ollama /api/embed endpoint, reducing API calls from N to N/10 for faster uploads
- Multi-file uploads now automatically use ZIP compression (threshold: >1 file or >5MB total)
Fixed
- Silenced verbose httpx/httpcore logs that spammed 60+ lines per file upload
- Silenced Tika startup warnings ("Failed to see startup log message; retrying...")
- Status indicators (Ollama/Qdrant/Authenticated) now wrap responsively on mobile view
Docker Images
ghcr.io/strawberry-code/ragify:1.2.0-tikaghcr.io/strawberry-code/ragify:latest-tika
v1.1.4
What's Changed
Added
- Multi-arch Docker images support (linux/amd64 + linux/arm64) for Ubuntu and Mac compatibility
- Updated
/buildand/push-ghcrslash commands for multi-arch manifest workflow
Fixed
- Dockerfile.tika Tika pre-download script now uses heredoc and handles versioned JAR names
- Removed obsolete mcp_server.py reference from Dockerfiles
Docker Images (Multi-Arch)
ghcr.io/strawberry-code/ragify:1.1.4-tika(linux/amd64, linux/arm64)ghcr.io/strawberry-code/ragify:latest-tika(linux/amd64, linux/arm64)
Docker will automatically select the correct architecture for your platform.
v1.1.3
What's Changed
Fixed
- Frontend API calls now include credentials for proper session authentication
- Ollama context size reduced from 8192 to 2048 to match nomic-embed-text model limit
- Centralized logging configuration for consistent pipeline logs visibility
- UI now shows actual error messages instead of generic failure messages
- Tika JAR detection now supports versioned filenames and TIKA_JAR_PATH env var
- Improved Tika pre-download in Dockerfile.tika with verification and copy to expected path
Docker Images
ghcr.io/strawberry-code/ragify:1.1.3-tikaghcr.io/strawberry-code/ragify:latest-tika