MichaelAnthony/ragworkspace-local-ai-starter-pack

RAGWorkspace Local AI Starter Pack v3

This repository distributes a version-pinned convenience ZIP for the RAGWorkspace llama.cpp installer. Version 3 contains seven compact text/chat models and one text embedding model. RAGWorkspace packages the files but is not the author or owner of the included models.

All model credit, immutable upstream revisions, exact hashes, licenses, and original-model lineage are recorded below, in NOTICE.md, and in SOURCE-MANIFEST.json.

Download and integrity

File	Bytes	SHA-256
ragworkspace-llama-models-starter-v3.zip	13,271,316,345	002744fe2cd83219748d7903f0a0930dc962ae04cbb0686ea2051ea54d1fe8f6

The archive is approximately 12.36 GiB. During online installation, the verified archive and extracted model payload can temporarily coexist, so the app requires about 25.2 GiB free including safety headroom. It calculates the exact requirement from manifest.json and any resumable partial download.

RAGWorkspace checks the archive's exact byte length and SHA-256 before extraction, rejects unexpected or unsafe ZIP members, and verifies each GGUF's size, magic header, and SHA-256 while extracting it.

Included models

Role	Packaged GGUF	Bytes	Quantization	Pinned GGUF source	Original-model lineage	License
Chat, default	Bonsai-1.7B-Q1_0.gguf	248,302,272	Q1_0	prism-ml/Bonsai-1.7B-gguf@210a9e9	prism-ml/Bonsai-1.7B-unpacked@a7f720b	Apache-2.0
Chat	Qwen3.5-0.8B-UD-Q8_K_XL.gguf	1,186,443,520	UD-Q8_K_XL	unsloth/Qwen3.5-0.8B-GGUF@6ab4614	Qwen/Qwen3.5-0.8B@2fc0636	Apache-2.0
Chat	Qwen3.5-2B-UD-Q5_K_XL.gguf	1,466,687,744	UD-Q5_K_XL	unsloth/Qwen3.5-2B-GGUF@f6d5376	Qwen/Qwen3.5-2B@15852e8	Apache-2.0
Chat	Qwen3-1.7B-UD-Q5_K_XL.gguf	1,262,991,936	UD-Q5_K_XL	unsloth/Qwen3-1.7B-GGUF@d7f544e	Qwen/Qwen3-1.7B@70d244c	Apache-2.0
Chat	ibm-granite_granite-3.2-2b-instruct-Q6_K.gguf	2,080,561,792	Q6_K	bartowski/ibm-granite_granite-3.2-2b-instruct-GGUF@9be2c10	ibm-granite/granite-3.2-2b-instruct@641593c	Apache-2.0
Chat	Ministral-3-3B-Instruct-2512-UD-Q5_K_XL.gguf	2,479,158,560	UD-Q5_K_XL	unsloth/Ministral-3-3B-Instruct-2512-GGUF@7564922	mistralai/Ministral-3-3B-Instruct-2512@b35d4df	Apache-2.0
Chat	gemma-4-E2B_q4_0-it.gguf	3,349,516,256	Q4_0 QAT	google/gemma-4-E2B-it-qat-q4_0-gguf@675cff4	google/gemma-4-E2B-it-qat-q4_0-unquantized@d884c6a	Apache-2.0
Embedding	Qwen3-Embedding-0.6B-f16.gguf	1,197,629,056	FP16	kalle07/embedder_collection@2ed1b74	Qwen/Qwen3-Embedding-0.6B@97b0c61	Apache-2.0

The exact per-model SHA-256 values are in manifest.json and SOURCE-MANIFEST.json.

Bonsai filename note

The upstream repository currently contains both Bonsai-1.7B.gguf and Bonsai-1.7B-Q1_0.gguf. They have the same 248,302,272-byte length and the same SHA-256 (3d7c6c90…cae5f7e3), so they are aliases rather than different builds. This pack includes only the descriptive Q1_0 filename. The file is about 248 MB, not 250 KB.

RAGWorkspace installation and live catalog

1. Install the managed llama.cpp runtime in RAGWorkspace.
2. Open Native llama.cpp settings.
3. If this release was just published, select Check Hugging Face for updates to bypass the validated one-hour catalog cache.
4. Select Download & install model pack.
5. Keep RAGWorkspace open while it downloads, verifies, extracts, and registers the models.

The app reads the small catalog from this repository's main/manifest.json, but that manifest may reference package archives only inside this same repository at an immutable release tag or 40-character commit. If the live catalog cannot be fetched or validated, the app uses its bundled manifest or last validated cache. Downloads and upgrades are always user-initiated.

Control of this Hugging Face repository is the catalog's publisher-authentication boundary. SHA-256 verifies that the downloaded bytes match the catalog, but the catalog supplies that digest and is not independently signed. Release policy treats published vN tags as immutable. The app also pins the exact ID, version, and SHA shown at confirmation time, so a later catalog change requires the user to review and confirm again. The embedded app manifest is a minimum-version floor; publisher-directed rollback remains possible to a release at or above that floor.

Bonsai is the default chat model. Qwen3 Embedding is registered only for RAG and embedding use, with explicit last-token pooling, and is never selected as a chat model.

For offline installation, put manifest.json and the complete ZIP in a native_model_assets folder beside the RAGWorkspace installer or portable EXE, then select Install model pack.

Text-only scope

This package supports text/chat generation and text embedding. It intentionally omits the optional Gemma, Qwen3.5, and Ministral multimodal mmproj files, so the pack alone does not enable image, audio, or video prompting.

License, provenance, and safety

The pinned source/original-model lineage records Apache-2.0 for all included weights. kalle07/embedder_collection does not itself declare license metadata, so the Qwen3 embedder's Apache-2.0 attribution follows its pinned original Qwen/Qwen3-Embedding-0.6B repository. A complete Apache License 2.0 text is provided beside and inside the ZIP. Qwen3.5's distinct license file, including its 2026 Alibaba Cloud copyright notice, is preserved as LICENSE-QWEN3.5-APACHE-2.0.txt; the upstream Bonsai notice is preserved as NOTICE-BONSAI.txt.

The model weights are redistributed byte-for-byte from the pinned GGUF sources; RAGWorkspace did not further quantize or fine-tune them. Collection-level changes are limited to assembling a deterministic ZIP64 archive, adding legal and provenance material, and declaring runtime roles and selection policy.

Model output can be inaccurate, offensive, unsafe, or unlawful. Users remain responsible for evaluating output, respecting third-party rights, and complying with applicable law and each model's license. This summary is informational, does not replace the license or upstream model cards, and is not legal advice.

Version 3 replaces the v2 chat lineup, including both LFM2.5 1.2B files and the uncensored Gemma E4B derivative, while retaining the Qwen3 RAG embedder. Existing v1, v2, and v2.1 tags remain immutable historical releases.