AI Flow

For Fable: John now wants early EDA library assembly plus a full component 3D tour and a separate 5-second overview for compose. I have rebuilt and tested the components/provenance command (PR #4), and am building the tour renderer, but need the actual released source to integrate its overview into compose without losing existing features.

Installed version: adom-aiflow 0.1.25

Published source read directly from adom/adom-aiflow: crates/adom-aiflow/src/main.rs, flows/board.json, workspace crates. The main.rs command enum lacks installed compose, tour, recut, prompt and done commands. The workspace Cargo version is 0.1.0. This is not a compiler failure: the supplied source builds, but it is not the current release command surface. Please publish the complete release source or point to the correct maintained repo/branch.

Installed help:

The board loop as one command: placement, routing, pours, gate, landing, measurement, analysis, finish line

Usage: adom-aiflow [OPTIONS] <COMMAND>

Commands:
  start     Start a run: copies the board, stamps the prompt time
  plan      The stages this board needs and who can take each
  take      Record who takes a stage: route=ai pour=binary
  stage     Stamp a stage you do yourself: start or end
  place     Placement helpers: pack (--wish), check (--moves), land (--moves)
  route     The grid router on the current board
  pour      Pours, keepouts and vias from the spec (and the routing plan for Kelvin keepouts)
  gate      KiCad DRC offline on the routing plan
  land      Replay live through the bridge: route, vias, pours, moves
  measure   Filled copper per layer from the live board
  analyze   current or thermal
  finish    The finish line; refuses until every check passes
  step      Declare the step you are working on (placement, routing, pours, current, thermal, capture, finish, or your own). --back marks a return to an earlier step; --why says what sent you back
  deliver   The end of the run as the human sees it: the moment you say "done, here is your video". Needs finish first and the video file
  report    The run's own README (the sub-README): steps, times, returns, clips and screenshots so far; --push puts it on a wiki page's repo under docs/runs/<id>/, --refresh reloads it in the human's browser. Runs by itself at every step change and at deliver once a page is set
  giveback  Record an issue or PR your AI filed back on adom/adom-aiflow (or any page) during this run; deliver counts them
  recut     Re-cut every clip of the run (the 10x cut and the motion-only action cut with the real-time clock), for runs recorded before a cut existed or after the cut rules improved
  compose   Compose the run's final video from the raw clips: only the frames where something moved, in order, paced so the whole video fits --max-seconds (2 minutes by default); the AI's analysis drawings as fullscreen shots; the narration cut to fit each segment, never the other way round. Replaces the delivered video when the run is already delivered
  artifact  Register a file the AI made during the current step (an analysis drawing, a plot): it goes on the run page under the step and, for images, into the final video as a fullscreen shot at that point of the step
  usage     Snapshot the engine's plan usage (percent used per limit window) into the ledger now; the binary also does this at start, every step, finish and deliver
  models    Check every footprint's 3D model on the live board (kicad_model_check): the models step. ERROR lists what is unresolved so the AI can fetch or build the models before placement
  tour      Walkthroughs on camera: `tour 3d` opens the 3D Viewer and runs the scripted tour as this step's clip; `tour nets` selects each key net (loads, wide nets, Kelvin) so the editor highlights its pours and traces together, frames it, and marks the clip
  ledger    Print the append-only ledger (run.jsonl): every turn with its thinking and tool seconds, every step visit and return, every stage and outcome, in order
  capture   open, start, mark <label>, stop
  compare   The comparison page: a sub-README for this run and another (`compose --with` first), the side-by-side video first, then the numbers from both ledgers, the AIs' drawings side by side and the links to both run pages; --push puts it on the page under docs/comparisons/<a>-vs-<b>/
  status    Where the run stands
  prompt    A new human prompt on this run (a follow-up: "add the walkthroughs", "recut the video"): the AI's time for it counts from here to its `done`. --at is the paste time when you start late
  done      The AI's answer to the current prompt is complete ("done, here is ..."): the session's clock stops here. `deliver` is the done of the first task
  sessions  The run as sessions: each human prompt, the AI's active time on it, the idle gaps cut out (a gap over --idle-minutes with no command running is the human away, not the AI thinking)
  help      Print this message or the help of the given subcommand(s)

Options:
      --ai-thread <AI_THREAD>  The name of the AI conversation calling (required for state-changing commands)
      --run <RUN>              Run directory (default ./aiflow-run or $AIFLOW_RUN)
  -h, --help                   Print help
  -V, --version                Print version

Rebuilt source help:

The board loop as one command: placement, routing, pours, gate, landing, measurement, analysis, finish line

Usage: adom-aiflow [OPTIONS] <COMMAND>

Commands:
  start       Start a run: copies the board, stamps the prompt time
  components  Inventory every board reference for the early component quality review; optional MPN marking is an explicit preference
  plan        The stages this board needs and who can take each
  take        Record who takes a stage: route=ai pour=binary
  stage       Stamp a stage you do yourself: start or end
  place       Placement helpers: pack (--wish), check (--moves), land (--moves)
  route       The grid router on the current board
  pour        Pours, keepouts and vias from the spec (and the routing plan for Kelvin keepouts)
  gate        KiCad DRC offline on the routing plan
  land        Replay live through the bridge: route, vias, pours, moves
  measure     Filled copper per layer from the live board
  analyze     current or thermal
  finish      The finish line; refuses until every check passes
  step        Declare the step you are working on (placement, routing, pours, current, thermal, capture, finish, or your own). --back marks a return to an earlier step; --why says what sent you back
  deliver     The end of the run as the human sees it: the moment you say "done, here is your video". Needs finish first and the video file
  report      The run's own README (the sub-README): steps, times, returns, clips and screenshots so far; --push puts it on a wiki page's repo under docs/runs/<id>/, --refresh reloads it in the human's browser. Runs by itself at every step change and at deliver once a page is set
  giveback    Record an issue or PR your AI filed back on adom/adom-aiflow (or any page) during this run; deliver counts them
  ledger      Print the append-only ledger (run.jsonl): every turn with its thinking and tool seconds, every step visit and return, every stage and outcome, in order
  capture     open, start, mark <label>, stop
  status      Where the run stands
  help        Print this message or the help of the given subcommand(s)

Options:
      --ai-thread <AI_THREAD>  The name of the AI conversation calling (required for state-changing commands)
      --run <RUN>              Run directory (default ./aiflow-run or $AIFLOW_RUN)
  -h, --help                   Print help
  -V, --version                Print version

Workaround: test an isolated binary and additive tour helper; do not replace the installed production binary with this incomplete snapshot. I can continue the renderer and library manifest work while this is fixed.

Run: /home/adom/aiflow-esc-astra; source checkout: /home/adom/aiflow-esc-astra/aiflow-component-pr; board: /home/adom/aiflow-esc-astra/final-native/esc-g431-astra.kicad_pcb; spec: /home/adom/aiflow-esc-astra/spec.json.