Automation foundation model for tiny devices: 2-bit, 8-29 MB, tool calls, ASR, structured extraction and embeddings on phones, wearables, smart homes, robots, cars and microcontrollers.
fix: update engine version for Needle 3 to 3.2.0
HenryNdubuaku7faf26b09e7b
diff --git a/README.md b/README.mdindex 254d730..1c4bab7 100644--- a/README.md+++ b/README.md@@ -69,7 +69,7 @@ Word error rate on the full test splits, scored with the Whisper normalizers. Whneedle --model needle3.cact --model whistle.cact --tools tools.json --audio clip.wav```-One engine holds both models: `needle_load` reads whichever one a `.cact` carries, and `needle_complete` takes a clip wherever it takes text. It transcribes, answers the transcript against your tools, and returns one JSON object with the tool calls and the speech fields, the speech ones prefixed `audio_`. The transcription stays inside the engine, so audio in and tool calls out is one call.+One engine holds both models: `needle_load` reads whichever one a `.cact` carries, and `needle_complete` takes a clip wherever it takes text. It transcribes, answers the transcript against your tools, and returns one JSON object with the tool calls and the speech fields, the speech ones prefixed `audio_`. The transcription stays inside the engine, so audio in and tool calls out is one call, and it favours the option values your schemas enumerate, so a clip that names one is transcribed as the schema spells it.The benchmarks against Whisper and Moonshine, the architecture and the interactive demo are at [cactuscompute.com/whistle](https://cactuscompute.com/whistle); the weights and every platform engine are on [Hugging Face](https://huggingface.co/Cactus-Compute/whistle).diff --git a/needle/agent/fetch.py b/needle/agent/fetch.pyindex b7bc93f..583bfe9 100644--- a/needle/agent/fetch.py+++ b/needle/agent/fetch.py@@ -13,7 +13,7 @@ ENGINE_REPOS = {}ENGINE_VERSIONS = {2: "2.0.4",- 3: "3.1.0",+ 3: "3.2.0",WHISTLE: "2.0.0",}