Local Agent Studio 0.5.2 Alpha

Mixed-provider workflows and flexible email automation

Local Agent Studio 0.5.2 makes mixed local/cloud workflows clearer and adds a
narrow, explicit automatic-send option for email nodes.

Added and fixed

• Connect Gemini, llama.cpp, Ollama, and LM Studio agents in the same visual
  workflow. Agent selectors now show both provider and model.
• Choose Ask every time or Send automatically for this node on each Send
  Email function. Automatic mode displays a warning and does not weaken approval
  requirements for file writes, Python, MCP, or mutating HTTP.
• Gmail authentication failures now explain that SMTP requires a Google App
  Password rather than the normal Google password.
• Spaces pasted into Gmail App Passwords are normalized before secure storage.
• Provider discovery and first-run Gemini/Ollama fixes from 0.5.1 are included.

Verification

• 23 backend execution tests passed.
• 6 renderer tests passed.
• TypeScript production checks passed.
• The bundled backend detected 2 installed Ollama models and completed a real
  input-to-agent-to-output workflow with the expected FLOW_OK result.
• The packaged UI audit exercised workflow editing, the Send Email approval
  selector, mixed-provider agent choices, runs, settings, and onboarding.
• Word and Excel creation completed against the bundled backend.
• Email approval and automatic-send paths were executed against a controlled
  SMTP test implementation. Real delivery requires valid credentials supplied
  by the user and was not claimed as part of this release test.

SHA-256 (Local-Agent-Studio-0.5.2-Setup.exe):
362F106E54C231AF3022BE0653251019CBDEA0524BA74EEF05D1765607C52FFC

Alpha notice: this installer is not code-signed yet. Windows SmartScreen
may show an unknown-publisher warning. Verify the checksum, review workflow
permissions, and keep independent backups of important data.

Local Agent Studio 0.5.1

First-run provider reliability fix

This patch fixes the onboarding defect that could hide Gemini and incorrectly
block working Ollama or LM Studio models after a stale health check.

Fixed

• All four provider choices are always shown: Gemini, built-in llama.cpp,
  Ollama, and LM Studio.
• The Gemini choice opens a tested three-step Google AI Studio/API-key guide.
• Successfully discovered Ollama or LM Studio models remain selectable even if
  an earlier health probe timed out.
• Ollama and LM Studio now distinguish "installed but server stopped" from
  "not installed."
• A failure in one provider cannot hide the other providers or their models.

Verification

• 21 backend tests passed.
• 6 renderer tests passed.
• TypeScript production checks passed.
• Packaged backend returned all four providers and detected 2 live Ollama
  models on the validation PC.
• Packaged UI audit clicked the Gemini provider and verified the complete guide,
  API-key field, and Google AI Studio link.
• Packaged Word/Excel workflow smoke test completed successfully.

SHA-256 (Local-Agent-Studio-0.5.1-Setup.exe):
3A86F3CEA7984CAD2AA9FB5609734D8203A2443B1E118DCE8B744BCC5386E053

This prerelease installer is not code-signed yet, so Windows SmartScreen may
show an unknown-publisher warning. Verify the checksum before running it.

Local Agent Studio 0.5.0 alpha

Private early-alpha Windows build with the fast-start bundled backend, local and Gemini providers, visual workflows, approvals, file/image attachments, Word and Excel tools, email, per-agent capability controls, optional Python functions, and local stdio MCP integration. Validation passed: 19 backend tests, 3 renderer tests, TypeScript checks, frozen-backend workflow/Word/Excel smoke test, and deterministic packaged Electron UI audit across Home, Models, Agents, agent permissions, workflow editing, attachments, functions, Runs, Resources, Settings, and Onboarding in 29.4 seconds. The installer remains unsigned and may trigger Microsoft SmartScreen. SHA-256: F0B7B3298FA7B5B324002D9BDB7C0FE25735A02823E6FF48BA6344E9B76D0EA7