Timeline
- (Context) Defined a new capstone flow: unlock once, then complete a guided tour by directing Gemma in a single terminal session.
- (Context) Set hard requirements for local runtime services (Ollama + Docker) so the capstone consistently demonstrates “local-first agents + safe execution”.
- (Action) Refactored Module 09 into an enforced interactive guided tour inside Gemma’s shell.
- (Action) Added step gating: the tour only advances when the learner successfully runs the required tool step and then types
continue. - (Action) Re-authored the tour content into phases that introduce tools in a deliberate order, including “Try it now.” and “Choose one now.” moments.
- (Action) Introduced a dedicated demo workspace for the capstone so each tool has realistic files to operate on.
- (Action) Updated reset behavior to restore the demo workspace from a clean snapshot so learners can experiment without fear.
- (Action) Added launcher-level readiness checks that automatically bring up required services and verify the required local model is available.
- (Observation) Fixed a wrapper path-resolution bug that could cause tools to fail when their working directory differs from the module root.
- (Observation) Standardized tool execution to consistently produce structured output for gating, logging, and predictable learner feedback.
- (Open Thread) Validate service auto-start behavior across fresh beginner installs (Windows/macOS/Linux) and define the best fallback messaging when auto-start fails.
Context
- Module 09 is intended to feel “plug-and-play” for true beginners: one entry point, one terminal, clear instructions, and no guessing.
- The capstone needs to demonstrate the full course arc: build tools, run them safely, and coordinate them through an agent loop.
- The tour must be deterministic: learners see exact commands, execute one, then advance only after success.
Actions
- Refactored Module 09 into an enforced interactive guided tour inside Gemma’s shell.
- Implemented a tour engine that records the last successful tool run and uses explicit completion rules to gate progress.
- Reworked the tour script to present exact copy/paste commands for each tool step and require “hands-on” execution before moving on.
- Added a demo workspace with intentionally messy inputs (text, CSV, mixed files) plus safe sandbox targets for the Pilot step.
- Implemented “restore demo workspace” behavior in the module reset flow via a clean snapshot, preserving learner work in a safe holding area.
- Added automatic service readiness checks: ensure Ollama is reachable, ensure the required local model is installed, and ensure Docker is available for sandbox runs.
- Hardened tool wrappers so relative paths behave predictably and so tool output is consistent for tour gating and learner clarity.
Observations
- Enforced “do the step, then continue” removes common beginner failure modes (skipping steps, running commands in the wrong place, or missing prerequisites).
- A dedicated demo workspace is essential: it makes tool outcomes visible, repeatable, and easy to reset.
- Service readiness checks reduce confusion but still need robust cross-platform fallbacks, especially around Docker engine startup.
Open Threads
- Confirm the best cross-platform strategy for automatically starting Docker and Ollama, and define the exact learner-facing messages when auto-start is unavailable.
- Decide whether the web summarization step should include an explicit offline-safe alternative for environments without internet access.
- Consider whether the organize step should require an actual “move files” run (writes) or allow a dry-run completion for maximum safety.
Boundary Reminder: Seeds. No maintenance. No roadmap.