Skip to content

release: v7.12.0 — Ollama support for agents - #576

Merged
ArmenSl merged 6 commits into
masterfrom
development
Jul 20, 2026
Merged

release: v7.12.0 — Ollama support for agents#576
ArmenSl merged 6 commits into
masterfrom
development

Conversation

@ArmenSl

@ArmenSl ArmenSl commented Jul 20, 2026

Copy link
Copy Markdown
Collaborator

Summary

  • Ollama support for agents (new feature): agents can use locally-hosted Ollama models as the LLM provider and as the RAG embedding provider — wired through the WME agent-config UI, the BAF generator/templates, and the agent converters (LLMOllama). Docs + an agent-converter round-trip test added.
  • B-UML export round-trip (backend rider): the multi-object exporter now writes a single object-model banner, so re-importing a multi-object project yields exactly those models instead of extra empty ones (web editor Add association icons in the palette of the web mdoeling editor #161 follow-up).
  • BPMN pools/lanes (frontend rider, via submodule bump): lane/pool children move with their parent; pools/lanes preserved on assistant BPMN injection (Ends name and multiplicity shown when passing from an Binary association to a Inheritance #159).

Test plan

  • Agent converter round-trip incl. Ollama LLM + embedding — passing
  • Agent services + generators sweep — 417 passed
  • CI green
  • Full docs build: cd docs && make html — v7.12.0 page renders

prasadadhav and others added 6 commits July 3, 2026 09:19
The multi-object export (v7.11.2) wrote two '# OBJECT MODEL' headers per
model: project_builder's '# OBJECT MODEL N #' plus the builder's
'# OBJECT MODEL #' banner. The project importer's section splitter matched
both, and the import lines between them are non-empty, so re-importing the
generated file produced one real object diagram PLUS one spurious empty one
per object model.

object_model_to_code now writes a single numbered + titled banner
(# OBJECT MODEL N: "Name" #) and project_builder no longer emits its own
section header - the same 'builder owns the header' pattern already used for
NN models. Re-import yields exactly the object models present, with their
names preserved. Adds a round-trip regression test.

Fixes BESSER-Web-Modeling-Editor#161 (round-trip follow-up)
add Ollama support to BAF agent and RAG generation
…ix model typo

Follow-up polish for #565 (Ollama LLM support):
- Document the ollama provider + ollamaModels in agent_personalization.rst,
  including the local base_url/model config and Ollama as a RAG embedding provider.
- Add an agent-converter round-trip test asserting the Ollama LLM and Ollama RAG
  embedding (provider + base_url + model) survive Agent -> code -> JSON -> Agent.
- Fix ollamaModels recommendation typo: gemma4:e2b -> gemma3n:e2b (no such
  gemma4 on Ollama), and drop the stray trailing comma.
- Bump setup.cfg 7.11.2 -> 7.12.0 and add v7.12.0 release notes.
- Frontend submodule bump 32a6adfb -> 360dcfa6: adds the Ollama BAF config UI
  (#157) and carries the BPMN pool/lane fixes (#159) that landed on develop.
- Backend: Ollama LLM + RAG support (#565) plus the #161 B-UML export
  round-trip fix, and follow-up docs/test polish (ac450bf).
@ArmenSl
ArmenSl merged commit 2310e6d into master Jul 20, 2026
6 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants