Skip to content

Latest commit

 

History

History
187 lines (120 loc) · 7.22 KB

ROADMAP.md

File metadata and controls

187 lines (120 loc) · 7.22 KB

Ollama Workbench Update Roadmap

Based on a thorough code audit conducted February 2026. Prioritized by impact: fix broken things first, then eliminate duplication, then improve architecture, then add features.

Status: ALL PHASES COMPLETE (February 2026)

Phase 1: Stop the Bleeding ✅ COMPLETE

Commit 36bc02d — Runtime bugs and dead code that actively cause failures or confusion.

1.1 Fix Broken Provider Calls ✅

  • Fixed call_groq_api signature to match how callers actually use it.
  • Rewrote build.py provider calls to use canonical functions from openai_utils.py and groq_utils.py.
  • Removed wildcard imports and shadowing local function definitions from build.py.
  • Fixed total_tokens type collision between Chat and Multi-Model Chat.

1.2 Remove Dead Code ✅

  • Deleted fixed_main.py, integrated_main.py, robust_ollama_utils.py, simplified_ollama_utils.py
  • Deleted all fix_*.py one-time scripts
  • Deleted fixed_chat_interface.py, launch_fixed_chat.py, multimodel_fix.py
  • Removed dead navigation branches and duplicate elif blocks in main.py
  • Deleted modern_chat_interface.py, simple_modern_interface.py, and .bak copies
  • Deleted cruft files (MODERNIZATION_STATUS.md, test_report.json, etc.)
  • Updated .gitignore

1.3 Fix Dependencies ✅

  • Removed conflicting autogen line from requirements.txt
  • Removed unused streamlit-flow packages
  • Removed stub bs4 and redundant plotly-express
  • Moved dev tools to requirements-dev.txt
  • Updated stale model lists for Groq and OpenAI

1.4 Security Quick Fixes ✅

  • Removed /openai-key Flask endpoint that exposed plaintext API keys
  • Added os.chmod(API_KEYS_FILE, 0o600) after writing api_keys.json
  • Removed native_host_manifest.json writer from page load

Phase 2: Consolidate Duplication ✅ COMPLETE

Commit 950b38f — Created single sources of truth for duplicated functions.

2.1 Single load_api_keys / save_api_keys ✅

  • Canonical implementation in ollama_utils.py
  • Replaced all 10 duplicate definitions with imports

2.2 Single initialize_session_state ✅

  • session_utils.py:initialize_session_state() is the single source of truth
  • Removed 8+ duplicate versions across modules

2.3 Single call_ollama_endpoint ✅

  • Canonical version in ollama_utils.py
  • Removed copies from repo_docs.py and dead files

2.4 Unify Model Selection Variable ✅

  • selected_model is the only variable used everywhere
  • Deleted current_model and all synchronization code

2.5 Dissolve ui_elements.py ✅

  • Changed main.py to import directly from source modules
  • Deleted ui_elements.py

2.6 Fix Logging ✅

  • Single logging.basicConfig() call in main.py
  • All other modules use only logger = logging.getLogger(__name__)
  • Reduced verbose session_utils logging to DEBUG

Phase 3: Architecture Improvements ✅ COMPLETE

Commit d93e433 — Structural changes for long-term maintainability.

3.1 Provider Abstraction Layer ✅

  • Created providers/base.py with BaseProvider ABC and ProviderResponse dataclass
  • Created providers/provider_ollama.py, provider_openai.py, provider_groq.py, provider_mistral.py
  • Added get_provider(name) factory function
  • Migrated multimodel_chat.py and other callers

3.2 Decouple Backend From Streamlit ✅

  • Extracted AI workflow logic into pure functions
  • Kept Streamlit UI as thin wrappers

3.3 Fix the Chat Interface Stack ✅

  • Merged fixed_chat_settings.py fixes into session_utils.py
  • Removed monkey-patching in enhanced_chat_interface.py
  • enhanced_chat_interface.py is now a thin wrapper

3.4 Stop Network Calls on Every Rerun ✅

  • Cached model list with @st.cache_data(ttl=60)
  • Only refreshes on explicit user action

3.5 Remove the Embedded Flask Server ✅

  • Deleted Flask app, run_flask(), send_port_to_extension(), and native_host_manifest.json writer

Phase 4: Test and Stabilize ✅ COMPLETE

Commit d144e67 — Fixed broken tests and added coverage.

4.1 Fix Broken Tests ✅

  • Deleted 12 test files testing deleted/non-existent modules
  • Fixed test_chat_integration.py, test_ollama_utils.py, test_build.py imports
  • Updated 1,243 mock patch paths from old flat module names to new package paths (626 failures → 261)

4.2 Add Coverage for Untested Modules ✅

  • tests/test_tool_playground.py — 62 tests
  • tests/test_nodes_workflow.py — 44 tests
  • tests/test_collaborative_workspace_new.py — 45 tests
  • tests/test_model_management_new.py — 39 tests
  • Total: 190 new tests, all passing

4.3 Remove Hardcoded Test Logic From Production Code ✅

  • Removed if model == "llama2" and text == "This is a test text..." special case from chat_interface.py
  • ModelMemoryHandler.segment_text() now dispatches purely on model_type

Phase 5: Modernize ✅ COMPLETE

Commit cb2ca96 — Dynamic model lists, CI/CD, and UX improvements.

5.1 Update Provider Model Lists Dynamically ✅

  • get_openai_models(), get_groq_models(), get_mistral_models() fetch from APIs with @st.cache_data(ttl=300)
  • Automatic fallback to hardcoded lists when API keys are missing or calls fail
  • Lazy wrapper functions in ollama_utils.py to avoid circular imports
  • Updated 17 consumer files to use dynamic functions instead of static constants

5.2 Migrate to a Package Structure ✅

  • Created ollama_workbench/ package with 8 sub-packages: providers/, chat/, workflows/, knowledge/, models/, server/, core/, ui/
  • Moved 128 root .py files into proper package structure (commit 19ec3b7)
  • Updated all imports

5.3 Replace ChromaDB With a Maintained Alternative ✅

  • Evaluated: chroma_client.py is dead code — nothing imports it
  • Annotated as unused; ChromaDB dependency retained for potential future use
  • RAG features use other vector storage mechanisms

5.4 Add CI/CD ✅

  • GitHub Actions workflow: ruff check + pytest
  • ruff.toml configuration (Python 3.12, line-length 120)
  • requirements-dev.txt for dev dependencies

5.5 Streamline the Chat Experience ✅

  • Voice Chat nav item only shown when pyaudio dependencies are available (no more error pages)
  • Dynamic nav building based on available features

Summary

Phase Status Commit Description
1. Stop the Bleeding ✅ Complete 36bc02d Fixed runtime crashes, removed ~30 dead files, fixed dependencies
2. Consolidate Duplication ✅ Complete 950b38f Eliminated 10x load_api_keys, 8x initialize_session_state, 5x call_ollama_endpoint
3. Architecture Improvements ✅ Complete d93e433 Provider abstraction, Streamlit decoupling, chat stack cleanup
4. Test and Stabilize ✅ Complete d144e67 Fixed broken tests, added 190 new tests for critical modules
5. Modernize ✅ Complete cb2ca96 Package structure, CI/CD, dynamic model lists, conditional nav

All 5 phases completed February 2026.