935 Commits

Author	SHA1	Message	Date
Thorsten Sommer	08b6a190a8	Fix the verify job on a fresh checkout	2026-09-13 12:31:10 +02:00
Thorsten Sommer	ed17f703b3	Updated changelog	2026-09-13 12:22:07 +02:00
Thorsten Sommer	58b9aa42f9	Add a quality gate for tests, Clippy, and the model sources	2026-09-13 11:42:10 +02:00
Thorsten Sommer	caf2aba4b9	Read the context window from four more model lists	2026-09-13 10:59:41 +02:00
Thorsten Sommer	1c03a46e2f	Read the context window from a provider's model list	2026-09-13 10:28:10 +02:00
Thorsten Sommer	d8a30c4e60	Add the model limits to the expert settings	2026-09-13 08:54:44 +02:00
Thorsten Sommer	bb22b3c10a	Let a provider state its own window and image limits	2026-09-12 20:50:38 +02:00
Thorsten Sommer	8fdef4aa6c	Let the snapshot cover every model of the corpus	2026-09-12 20:33:13 +02:00
Thorsten Sommer	672237a225	Tell embedding providers which tokenizer their model uses	2026-09-12 20:16:27 +02:00
Thorsten Sommer	5326a0bf6c	Name the tokenizer a model uses, where its vendor does	2026-09-12 20:01:29 +02:00
Thorsten Sommer	2509fc09d1	Say when a chat carries more images than the model takes	2026-09-12 19:43:18 +02:00
Thorsten Sommer	4d43164577	Check image counts against the limits vendors state	2026-09-12 19:32:08 +02:00
Thorsten Sommer	8468cdc7be	Recompute the token count instead of announcing it	2026-09-12 18:56:20 +02:00
Thorsten Sommer	5ee4685f6f	Count the whole conversation against the model's window	2026-09-12 17:56:01 +02:00
Thorsten Sommer	e8ec2c1bef	State the context window where a vendor documents one	2026-09-12 15:53:03 +02:00
Thorsten Sommer	6d80e6de8a	Let organizations describe their own models	2026-09-12 15:29:55 +02:00
Thorsten Sommer	f4078856d1	Split the reasoning parameters into dialects	2026-09-12 14:36:56 +02:00
Thorsten Sommer	849f072f3e	Answer the model kind from the registry	2026-09-12 14:22:21 +02:00
Thorsten Sommer	51613e3748	Delete the old capability rules	2026-09-12 14:12:26 +02:00
Thorsten Sommer	b6defc8b4b	Keep non-chat models out of the chat lists	2026-09-12 13:52:10 +02:00
Thorsten Sommer	77130a450a	Read model capabilities from the profile everywhere	2026-09-12 13:36:39 +02:00
Thorsten Sommer	429ca8c739	Resolve model profiles through the registry	2026-09-12 10:50:12 +02:00
Thorsten Sommer	9307f2c363	Move the model kinds into the rule engine	2026-09-12 10:13:59 +02:00
Thorsten Sommer	77c95e6f41	Fixed warnings	2026-09-12 09:50:10 +02:00
Thorsten Sommer	7eec39c4a7	Port the remaining open weights we decided to name	2026-09-12 09:41:39 +02:00
Thorsten Sommer	ac547ccb79	Port the Gemma, Nemotron, and Phi families	2026-09-12 09:08:42 +02:00
Thorsten Sommer	ad3a25c80f	Port the Mistral open weights and the GLM family	2026-09-11 22:10:40 +02:00
Thorsten Sommer	9c65ab6c62	Port the Llama and Qwen open weights	2026-09-11 21:35:47 +02:00
Thorsten Sommer	402da2da96	Port the DeepSeek, Perplexity, and xAI families	2026-09-11 21:06:44 +02:00
Thorsten Sommer	5905f074b6	Port the Alibaba Cloud families and refuse ambiguous rule inheritance	2026-09-11 20:52:42 +02:00
Thorsten Sommer	31a1aec69c	Port the Mistral families, including the release date logic	2026-09-11 20:33:16 +02:00
Thorsten Sommer	430867089e	Spelling	2026-09-11 20:20:01 +02:00
Thorsten Sommer	c6cbce6e01	Port the Anthropic and Google families	2026-09-11 20:18:23 +02:00
Thorsten Sommer	730bd1c7c4	Port the OpenAI families and add the model registry	2026-09-11 20:01:45 +02:00
Thorsten Sommer	31b655cf3b	Add one model host per provider and the unwrapping walk	2026-09-11 19:28:06 +02:00
Thorsten Sommer	d3b28134d9	Add the model family API, its compile-time registry, and MWAIS0013	2026-09-11 18:28:24 +02:00
Thorsten Sommer	3302915d84	Add the model matching engine and make capabilities a flags enum	2026-09-11 17:50:36 +02:00
Thorsten Sommer	87566c74e4	Add a model corpus and characterization tests for the capability rules	2026-09-11 17:06:40 +02:00
Thorsten Sommer	de775048f0	Added a NUnit test project for the whole solution	2026-09-11 16:37:09 +02:00
Peer Hogeterp	d21e09dd1e	Updated the AI Studio logo (#957)	2026-09-11 15:27:19 +02:00
Thorsten Sommer	2834529753	Fixed model capabilities and how model names are matched (#958)	2026-09-11 15:24:02 +02:00
Thorsten Sommer	e64c51bd57	Moved the custom tokenizer into the expert settings and added a drop zone for it (#956)	2026-09-10 19:18:40 +02:00
Thorsten Sommer	41d788fc9b	Fixed the visual briefing scrolling and raise it to beta (#954)	2026-09-10 17:46:27 +02:00
Thorsten Sommer	eb65b16851	Deliver each drop to the zone under the cursor (#953)	2026-09-10 15:28:08 +02:00
Paul Koudelka	c7b42bee96	Added local RAG (#756) ... Co-authored-by: Thorsten Sommer <SommerEngineering@users.noreply.github.com>	2026-09-09 18:43:37 +02:00
Thorsten Sommer	aca4404351	Quiet secret store logs and fix untranslated tool messages (#950)	2026-09-05 19:38:54 +02:00
Thorsten Sommer	88bf475847	Added admin dialog for tool configuration exports (#949)	2026-09-05 19:16:54 +02:00
Thorsten Sommer	4141e7c504	Support multiple web search backends: SearXNG, Staan, and Tavily (#948)	2026-09-04 22:14:29 +02:00
Peer Hogeterp	4d8d30e15e	Added tool calling support (#731) ... Co-authored-by: krut_ni <nils.kruthoff@dlr.de> Co-authored-by: Thorsten Sommer <SommerEngineering@users.noreply.github.com>	2026-09-04 15:48:07 +02:00
Thorsten Sommer	b00c3f9ab3	Prepared release v26.8.2 (#946)	2026-08-31 10:00:09 +02:00