diff --git a/STATUS.md b/STATUS.md index 5a6a575..506388f 100644 --- a/STATUS.md +++ b/STATUS.md @@ -1,6 +1,6 @@ # BatLLM Status -Last updated: 2026-05-10 01:04 +Last updated: 2026-05-10 17:21 BatLLM is a Python/Kivy research, education, and game project for exploring AI-mediated play, prompt quality, LLM behaviour, and local-model workflows. The repository currently contains a playable local desktop game, a standalone read-only Game Analyzer, local Ollama lifecycle and model-management helpers routed through `modelito`, release-bundle tooling, Homebrew formula generation, generated API reference artefacts, and maintained user/developer documentation. @@ -86,8 +86,8 @@ This status update followed a repository-wide audit on 2026-05-09. The audit ins - Repository patch version was bumped from `0.3.5` to `0.3.6`. - Mock-Ollama smoke validation now accepts a responding `/api/version` endpoint when process-level inspection cannot identify the mock server as an Ollama process. - Generated Doxygen output under `docs/code/` was regenerated after source and version changes. Existing Doxygen warnings are mostly undocumented Kivy/test helper classes and remain non-blocking documentation debt. -- Architecture and runtime-flow SVG diagrams were redrawn with larger canvases, wrapped text, boundary-aligned connectors, plain arrow labels, and clearer label placement. -- `docs/index.html` is now the public GitHub Pages showcase site for `https://krahd.github.io/BatLLM/`; GitHub Pages now serves from branch `main` and path `/docs`, the hero no longer uses a diagram background or top-of-page gameplay image, the page uses a darker modernised visual system, grant/provenance notes are surfaced in a dedicated section, Tomas Laurenzo attribution remains in provenance/footer copy, the demo GIF is used once in-page, inline text links are no longer bolded, the hero eyebrow label has been removed, the hero right column shows the project logo at large scale, the screenshot has rectangular (zero-radius) corners, and documentation links in the hero and links grid point to the rendered GitHub pages for README, USER_GUIDE, CONTRIBUTING, and CREDITS. +- Architecture and runtime-flow SVG diagrams were redrawn with larger canvases, wrapped text, boundary-aligned connectors, plain arrow labels, and clearer label placement; the runtime-flow SVG canvas width now matches the architecture diagram width, right-side boxes were shifted apart for clearer spacing, then boxes 2A/2B and arrow labels were fine-tuned so right-side arrow labels sit cleanly over their connectors, including an additional small left shift of 2A/2B with matching arrow-anchor adjustments, followed by iterative width reductions of boxes 1, 2A, and 2B for improved separation. The architecture diagram was then adjusted so every box except `modelito` is 15% narrower, with connector anchors updated to the new box edges, and the Local Ollama Runtime box text was repositioned inside the box. +- `docs/index.html` is now the public GitHub Pages showcase site for `https://krahd.github.io/BatLLM/`; GitHub Pages now serves from branch `main` and path `/docs`, the hero no longer uses a diagram background or top-of-page gameplay image, the page uses a darker modernised visual system, grant/provenance notes are surfaced in a dedicated section, Tomas Laurenzo attribution remains in provenance/footer copy, the demo GIF is used once in-page, inline text links are no longer bolded, the hero eyebrow label has been removed, the hero right column shows the project logo at large scale, a stronger layered white glow is applied around the hero logo to soften visible border pixelation, the screenshot has rectangular (zero-radius) corners, and documentation links in the hero and links grid point to the rendered GitHub pages for README, USER_GUIDE, CONTRIBUTING, and CREDITS. - The git worktree was clean at the start of this audit; current changes are intentional audit/remediation updates. ## Current Implementation State @@ -275,4 +275,4 @@ The previous status report recorded these successful checks from the same releas - Design the 2.0 server contract before adding web or repository-backed prompt/game sharing. - Add broader tests for malformed model responses, slow startup, missing models, session compatibility, analyzer edge cases, and packaged first-run behaviour. -Last updated: 2026-05-10 01:04 +Last updated: 2026-05-10 17:21 diff --git a/docs/images/architecture-modelito.svg b/docs/images/architecture-modelito.svg index 0620584..d7ebe9f 100644 --- a/docs/images/architecture-modelito.svg +++ b/docs/images/architecture-modelito.svg @@ -25,7 +25,7 @@ BatLLM modelito integration architecture The Kivy tools share one configuration layer and route local-model work through the BatLLM LLM facade before delegating to modelito. - + Launchers run_batllm.py @@ -33,7 +33,7 @@ run_tests.py - + Config And State src/configs/config.yaml @@ -44,7 +44,7 @@ last_served_model - + Configurator Tooling src/configs/configurator.py @@ -52,7 +52,7 @@ console stream - + Gameplay Surface GameBoard builds prompt history @@ -61,7 +61,7 @@ saved sessions stay replayable - + Analyzer Surface analyzer_main.py replay UI @@ -69,7 +69,7 @@ prompt and response inspection - + BatLLM LLM Facade src/llm/service.py @@ -91,35 +91,35 @@ download_model_progress - - Local Ollama Runtime - - real local Ollama server - installed model set - warm, download, delete - serve, stop and start + + Local Ollama Runtime + + real local Ollama server + installed model set + warm, download, delete + serve, stop and start - - launch + + launch - + launch - - shared config + + shared config - - config-driven tooling + + config-driven tooling - - prompt lifecycle policy + + prompt lifecycle policy - - replay compatibility + + replay compatibility - - delegate + + delegate service calls @@ -127,6 +127,6 @@ status and CLI - - prompt/model ops + + prompt/model ops diff --git a/docs/images/request-flow-modelito.svg b/docs/images/request-flow-modelito.svg index 837293c..e423c2e 100644 --- a/docs/images/request-flow-modelito.svg +++ b/docs/images/request-flow-modelito.svg @@ -1,4 +1,4 @@ - + BatLLM local model request flow The path from selecting or downloading an Ollama model to gameplay prompts, provider responses, parsing, and replay storage. @@ -20,11 +20,11 @@ - + Local model request and replay flow Model selection, readiness checks, downloads, gameplay prompts, and saved sessions share one modelito-backed path. - + 1. Select Or Refresh A Model @@ -39,34 +39,34 @@ Local Model Available? - - 2A. Serve Existing Model - - save llm.model - resolve timeout through - src/llm/service.py - modelito readiness check - record last_served_model + + 2A. Serve Existing Model + + save llm.model + resolve timeout through + src/llm/service.py + modelito readiness check + record last_served_model - - 2B. Download Remote Model - - modelito download progress - update status and output log - refresh local list - model becomes playable + + 2B. Download Remote Model + + modelito download progress + update status and output log + refresh local list + model becomes playable - - Local Ollama - - real local server - installed models - warm and pull - delete safeguards - serve/stop/start - provider responses + + Local Ollama + + real local server + installed models + warm and pull + delete safeguards + serve/stop/start + provider responses @@ -89,40 +89,40 @@ returns assistant text - - 5. Parse And Persist - - gameplay response is parsed - into actions and narrative - HistoryManager stores - prompts and outcomes - analyzer replay stays valid + + 5. Parse And Persist + + gameplay response is parsed + into actions and narrative + HistoryManager stores + prompts and outcomes + analyzer replay stays valid - - refresh + + refresh - - local + + local - - download + + download - - warm/inspect + + warm/inspect - - pull model + + pull model - - response text + + response text prompt - - record outcome + + record outcome - + after model is ready diff --git a/docs/index.html b/docs/index.html index d47936a..ca9e911 100644 --- a/docs/index.html +++ b/docs/index.html @@ -200,6 +200,10 @@ height: 300px; object-fit: contain; opacity: 0.88; + filter: + drop-shadow(0 0 12px rgba(255, 255, 255, 0.5)) + drop-shadow(0 0 28px rgba(255, 255, 255, 0.36)) + drop-shadow(0 0 42px rgba(255, 255, 255, 0.22)); } h1 {