deepanwadhwa · Aug 30, 2026 · Aug 20, 2026 · Aug 28, 2026 · Aug 29, 2026 · Aug 29, 2026
diff --git a/.gitignore b/.gitignore
@@ -1,4 +1,5 @@
 build/
+/build-linux-ubuntu/
 /build-ci-*/
 *.qws
 *.dSYM/

diff --git a/CLAUDE.md b/CLAUDE.md
@@ -227,7 +227,6 @@ make            # portable build
 make omp        # multithreaded (brew install libomp first, macOS)
 make test       # self-contained: stubs engine + network, tiny fixtures, no 24 GB model
                 # covers the compiled gateway, Chutni, and the Kimi preflight
-make test-all   # the above plus the Gigatoken adapter (needs `cargo +nightly`)
 ```
 
 - Model (24 GB, hard-linked, never copy): `~/Documents/samosa-models/qwen36_group32_i8`

diff --git a/Makefile b/Makefile
diff --git a/NOTICE b/NOTICE
@@ -24,5 +24,21 @@ contributors, licensed under the MIT License. The encoder/decoder inference
 loop in src/samosa_summarizer.cpp is adapted from llama.cpp's `simple`
 example under that license.
 
+The optional VisionPsy-Nano 460M BF16 weights are the MLX conversion published
+by KaedeTai at KaedeTai/VisionPsy-Nano-460M-MLX, pinned by Samosa to revision
+ca7a1bad19a08fe8e474bee24275a2e377bbabba. The upstream model is
+qvac/VisionPsy-Nano-460M, licensed under the Apache License, Version 2.0.
+Samosa executes the conversion through its independent native MLX C++ adapter;
+the Python reference runtime is not distributed.
+
+The optional Molmo2 4B package is converted from allenai/Molmo2-4B revision
+042abfa7a38879a376cec03d949eff0aefaa0600, copyright the Allen Institute for
+AI and contributors, licensed under the Apache License, Version 2.0. Samosa's
+native conversion keeps the vision tower and connector in BF16 and converts
+large language matrices to affine Q4/group-64. The original FP32 checkpoint,
+Python reference runtime, Transformers, and PyTorch are not distributed by
+Samosa.
+
 Samosa Chat is not affiliated with or endorsed by the colibrì project,
-Alibaba Cloud, the Qwen team, Falcons.ai, or the llama.cpp project.
+Alibaba Cloud, the Qwen team, Falcons.ai, the llama.cpp project, qvac, KaedeTai,
+the VisionPsy project, AllenAI, or the Molmo project.
diff --git a/README.md b/README.md
@@ -20,8 +20,14 @@ Samosa is a local model application and HTTP gateway. Models can be downloaded,
 - Starts with zero installed models.
 - Includes a browser interface for downloading models from a validated catalog.
 - Supports resumable and verified model downloads.
+- Automatically routes chat images and PDFs through local text/OCR and an
+  on-demand visual specialist; oversized inputs are bounded from live
+  memory/thermal conditions and multi-page work has no arbitrary page cap.
+- Adds optional, on-demand Molmo2 4B intelligence as the preferred installed
+  image/video specialist without a Python runtime or always-resident second
+  model; VisionPsy remains the smaller compatibility fallback.
 - Provides an internet-search feature that does not require an API key.
-- Integrates Chutni folder memory via a native Gigatoken adapter for document indexing.
+- Integrates portable Chutni folder memory through the bundled local `chutni-mcp` service.
 
 ## Models
 
@@ -31,8 +37,22 @@ Samosa is a local model application and HTTP gateway. Models can be downloaded,
 | [Maple](https://huggingface.co/deepanwa/Samosa-Chat-Maple-2bit-SSD) | 5.4 GB | 2-bit MoE with routed experts streamed from SSD | Samosa Metal engine | MIT |
 | [Bonsai 27B 1-bit](https://huggingface.co/prism-ml/Bonsai-27B-gguf) | 3.8 GB | 1-bit GGUF reasoning model | Prism llama.cpp | Apache-2.0 |
 | [Ornith 1.0 9B](https://huggingface.co/deepreinforce-ai/Ornith-1.0-9B-GGUF) | 5.6 GB | Q4_K_M GGUF coding and reasoning model | Prism llama.cpp | MIT |
-
-Only one model is loaded in memory at a time. Selecting a new model unloads the active backend before launching the next.
+| [VisionPsy-Nano 460M](https://huggingface.co/KaedeTai/VisionPsy-Nano-460M-MLX) | 1.01 GB | Auxiliary vision specialist for chat attachments (macOS arm64) | Samosa MLX Metal C++ helper | Apache-2.0 |
+| [Molmo2 4B](https://huggingface.co/allenai/Molmo2-4B) | ≤4.0 GiB native package | Preferred on-demand image/video specialist (macOS arm64) | Samosa MLX/AVFoundation C++ helper | Apache-2.0 |
+
+Only one primary chat model is loaded in memory at a time. When the optional
+Molmo2 package is installed it handles admitted visual image and video work;
+VisionPsy-Nano 460M is the standard-image fallback when Molmo is absent. On
+constrained Macs, Samosa stops the primary model, runs one bounded Molmo
+operation, and unloads it. If a conversation starts with one image or video,
+Molmo's own text and point coordinates are returned directly while the selected
+text model restarts for later turns. Later/composite visual workflows may use a
+grounded, greedy, non-thinking synthesis pass. Molmo2 is a locally built,
+hash-verified package: the app never
+downloads its 19.4 GB FP32 source checkpoint. Samosa contains zero runtime
+Python dependencies. The real 16 GiB-machine conversion and zero-swap
+image/video qualification are recorded in
+[`docs/regressions/molmo2-4b-q4-2026-08-26.md`](docs/regressions/molmo2-4b-q4-2026-08-26.md).
 
 > **Credits.** The Qwen runtime is based on [colibrì](https://github.com/JustVugg/colibri) by JustVugg and runs the model released by [Qwen](https://huggingface.co/Qwen). Bonsai is released by [PrismML](https://huggingface.co/prism-ml), and Ornith by [DeepReinforce](https://huggingface.co/deepreinforce-ai). GGUF execution uses PrismML's [llama.cpp fork](https://github.com/PrismML-Eng/llama.cpp). Samosa is an independent project.
 
@@ -94,7 +114,9 @@ Model downloads:
 | Bonsai | 3.8 GB (+17 MB runtime) |
 | Ornith | 5.6 GB (+17 MB runtime) |
 | Qwen | 24.0 GB |
-| All Models | 33.4 GB (+17 MB runtime) |
+| VisionPsy (macOS arm64 only) | 1.01 GB |
+| Molmo2 native Q4 package (optional, macOS arm64) | ≤4.0 GiB |
+| All Models | 34.4 GB (+17 MB runtime) |
 
 ## App Usage
 
@@ -105,11 +127,33 @@ samosa app
 Access the interface at <http://127.0.0.1:8642>. Available settings include:
 
 - Model selection and downloading.
+- Grouped Text LLM, Vision, OCR, and Voice model settings with live readiness.
+- Automatic chat-attachment routing, hardware-adaptive image resizing, and
+  honest text/OCR-only partial answers if visual analysis fails.
+- Streamed MP4/MOV/M4V attachments with bounded overview, temporal, or
+  exhaustive Molmo2 coverage and explicit unprocessed intervals.
 - Context size configuration (`auto` or explicit limits).
 - Automatic and manual conversation compaction.
 - Inference parameters (thinking budget, max tokens, seed) and optional web context tools.
+- Opt-in Developer mode with one correlated local trace across routing, OCR,
+  VisionPsy, evidence assembly, and the final model response.
 
 State and conversation ledgers are persisted under `~/.samosa/chats`.
+Developer traces are off by default because they contain full prompts and
+document/model contents; see [docs/DEVELOPER_MODE.md](docs/DEVELOPER_MODE.md).
+
+To use the Mac's models from a phone or another computer on the same Wi-Fi:
+
+```sh
+samosa app --lan
+```
+
+Samosa opens the normal local app and prints a clean `Other devices` link.
+Open it on the other device and enter the LAN password (`password1234` for
+now). Keep the Mac awake, allow incoming connections if macOS asks, and stop
+sharing with `samosa serve --stop`. See the full
+[LAN access guide](docs/LAN_ACCESS.md) for the trust boundary, password
+sessions, shared-model concurrency, model switching, and troubleshooting.
 
 ### Folder memory with Chutni
 
@@ -203,7 +247,8 @@ Test logs and regression benchmarks are located in [docs/regressions](docs/regre
 ## Privacy & Security
 
 - All inference and chat data remain local to the host machine.
-- The gateway server binds exclusively to `127.0.0.1`.
+- The gateway binds exclusively to `127.0.0.1` unless the user explicitly
+  starts password-protected LAN access with `--lan`.
 - No telemetry or external accounts are used.
 - Downloads are limited to pinned Hugging Face repository assets and Prism GitHub release binaries.
 - External web search tools are opt-in.
@@ -220,6 +265,7 @@ Test logs and regression benchmarks are located in [docs/regressions](docs/regre
 |---|---|
 | Installation & Model Storage | [docs/INSTALL.md](docs/INSTALL.md) |
 | App & Terminal Usage | [docs/USAGE.md](docs/USAGE.md) |
+| Password-protected LAN access | [docs/LAN_ACCESS.md](docs/LAN_ACCESS.md) |
 | Model Specifications & Network | [docs/MODELS_AND_INTERNET.md](docs/MODELS_AND_INTERNET.md) |
 | Gateway API | [docs/SERVE_API.md](docs/SERVE_API.md) |
 | Chutni folder memory | [docs/CHUTNI_PROTOCOL.md](docs/CHUTNI_PROTOCOL.md) |