run_workflow(base_only) queues the first pass alone (base_only.js cuts each
queued prompt; save nodes become previews, the canvas stays). When it ends,
ai_bases.py shows the previews on a picks panel; the picked one is copied
into the output folder and its seed held. The next run takes the kept
picture where the rest takes the base's picture, so the base is not drawn
again; a base handed on as a latent (mid-way switch) is drawn again from
its seed and compared with the kept one. Drawn again from the same seed, a
base once came out grainy on a 6 GB card, hence the kept picture.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
AMB only uses: training goes to its own plugin. This takes back the
record of what works (api/ai_evidence.py), the training network and
rounds (api/ai_training.py, ui_ai_network*.js, the round form), the
Gates' record switch, the ways' order by record and the external_skill
field of log_outside that only fed the record.
Kept, as part of using AMB:
- the user's word under a finished run, now only told to the AI
(ai_run_control.set_verdict, ui_ai_verdict.js);
- the Gates' "Custom nodes: only the ways'" rule;
- a picks block left to the AI takes its recommended picture.
Reverts 73518dc, 67bf5b2, 6647b05 and 49d05cf, then puts those three
back with tests in test_ai_run_watch, test_ai_gates and test_ai_panels.
Full suite and structure check pass.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The AI workbench's side column gets a training network view: base models
on the left, ways, skills and node packs on the right, a line for each
pair that ran, coloured by how it turned out (the user's word first, else
the AI's, dashed) and as thick as it ran often. A node or line shows its
record, last pictures and the pairs not tried yet in a card kept in sight.
"Start a training round" picks a skill (the AI's prompts with its advice
and without) or the ways of one use, up to four models and the pictures
each, and makes a task for the next AI (api/ai_training.py). Each run that
ends fills the trial it fits by what it used; the AI is told the next trial
and, once a model's trials are done, to compare them in a picks panel with
its own pick marked recommended. The record keeps the AI's pick apart from
the user's, and a picks block left to the AI takes its pick.
Tests: tests/test_ai_training.py; full suite and structure check pass.
Checked on an isolated ComfyUI: a round started from the view, four real
runs matched to their trials with the next one told each time, the
comparison panel and the user's pick reached the record and the graph.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Each run the AI starts now counts for the ways it still has, the skills
fetched before it and the custom node packs of its nodes, by the base
model of its main model (api/ai_evidence.py). How it went comes from what
the user does anyway: a picture picked in a picks panel or kept in the
gallery, one passed over, or their word under the run (Turned out well /
Not quite, told to the AI as their message).
begin_task and get_toolbox give the AI the record in a few words, add_nodes
a way's or the packs' own, and add_nodes(use=...) picks the use's way that
did best on the canvas's model. The Ways panel shows each way's and skill's
record by model. The Gates window gets "Custom nodes: Any / Only the ways'",
which keeps the AI to the ways' nodes and the task's picks.
Tests: tests/test_ai_evidence.py; full suite and structure check pass.
Checked on an isolated ComfyUI: a real run was recorded with its model and
skill, the verdict reached the AI, and the ways-only rule refused a custom node.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- tags_from_picture: WD14 tagger (the source's model and thresholds) on a
picture the user handed; run_status now carries what nodes showed as words
(texts: a tagger's tags), with a summary and next step for a run that made
words but no pictures.
- Optional nodes: a block node marked optional goes when its pack is missing,
and a link "match|decode" takes the first one there. The four upscales end in
KJNodes' ColorMatch (mkl, the source's) when it is installed; the panel says
"better with" it otherwise.
- LanPaint's next step keeps the prompt about the whole picture; Follow a pose
says OpenPose misreads anime (use control_depth), as the source's author did.
- The directory has a size cap: past it, block names alone, so collected ways
never make begin_task long; shorter headings and few-word lines.
Live-tested on the isolated 8199 ComfyUI: LanPaint (area changed 14.6, outside
0.0, 82 s), tiled 2x (167 s), depth ControlNet (follows the layout), OpenPose
(misreads an anime picture), background removal (8 s), tags (166 s with the
model download), upscale with ColorMatch (73 s).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
In the user's real tasks the AI spent many calls finding its way: reading the
canvas, searching node types, looking at combos and the activity log. The
directory gives it what ARCHITECTURE.md gives the code: each tool offered,
under its stage of the work, with its title. The stages are run, canvas,
models, prompts, combos, pictures, voices, the library, and the task and the
user. Seeing it, the AI can go straight to the tool it needs.
The directory is built from action_catalog's stages and the tool specs, so a
new tool shows up by itself, and a test checks that every stage has a heading.
It lists only the tools offered: about 1,100 characters for the daily set and
2,000 for the full one. begin_task's answer carries it, and so does
get_situation's while there is no task. The server's instructions and
begin_task's note point to it.
retext, the helper that rewrites a result's text from its data, moves from
ai_task_runs to mcp_tools, so both modules share it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Waiting on runs was the biggest share of the AI's calls in the user's real
tasks (85 of 316 run_status calls, about ten per ten-minute run). A new Gates
setting, "Waiting for runs", defaults to "Tell AI when done" (ai_gates.run_wait,
stored beside the tool set in ai_tools.json).
In that mode:
- run_workflow still waits its 15 s, so a short run comes back done with its
pictures.
- Otherwise the answer's note, `next` and the your_run sentence tell the AI to
say the run is going and end its turn, not to wait with run_status.
run_status still honours wait_seconds, so a model that polls anyway is no
worse off than before.
When the run ends before the AI knows, the server marks the step `call_ai`. The
workbench then shows lines to copy into the AI's chat (ui_ai_call.js): "It's
done, take a look", "Again with another seed", "Go on", or "It failed, find out
why". The mark goes as soon as the AI is told, or when a newer run starts, and
the lines go with it. The AI's next call brings the result and the pictures.
"AI waits" keeps the old behaviour. The setting takes effect at once, because
the words are made for each answer.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
In the user's real tasks, nearly every run ended with get_image_info, then
show_panel, then report: one extra call just to fetch the picture. The answer
that first tells the AI its run ended now carries up to four thumbnails of the
run's pictures (pictures_attached). That answer can be run_status, run_workflow's
own wait, or any other tool's answer.
It happens once per run. status() and news() mark the run when they first tell
its end, and show_pictures, called for every answer in mcp_server, attaches the
pictures to the marked run and clears the mark. Runs told of before this change
carry no mark, so they never bring old pictures along. The done `next` now says
to report with the pictures and to use a panel only to compare or ask. tell()
and show_pictures share _retext.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A small model plans poorly but follows a line well. begin_task,
run_workflow and run_status, add_node, connect_nodes and remove_nodes,
and an answered ask_user now carry next: the one thing to do now (wait
again, look at the pictures and report, wire what is still unwired, find
the missing model). The run's per-state note becomes that next.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A progress block on an AI panel is the AI's own number and never moves.
An AI drew one at 12% during a run, came back after the run with a new
panel and left the old bar standing, which read as live progress. Once
the task goes past a bar without an update (a run ends, a later panel or
note, the task ends), the stage now draws it grey and says when it was
written; update_panel records when a panel's blocks changed. The tool
text tells the AI a run's real progress shows already.
run_status before any run answers state "none" instead of failing, so a
task no longer shows a failed step for asking.
In a wide window the task column keeps its content to about 880 px in
its middle, as chat apps do, instead of stretching across the window.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
In one hour Gemini ran 63 shell commands next to 23 AMB calls: checking files, cropping and
comparing parts of pictures, running OpenPose itself to see a skeleton, and painting pixels
with OpenCV (it overwrote an output picture six times).
- api/mcp_look.py zoom_image: part of a picture in the output, input or temp folder shown
larger, the same part of a second picture beside it; read only.
- get_image_info reads the temp folder too; run_status lists a run's PreviewImage pictures
(ai_task_runs.outputs_of(record, "temp"), kept as the step's previews, not its results).
- begin_task and the repaint note: AMB's tools cover looking; change pictures only by running
workflows, never by editing picture files with code of one's own.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Antigravity saves a tool result over about 4,000 characters to a file and has Gemini open it;
in one evening 16 of its 32 file reads were AMB results (describe_canvas 9 times). Results are
now compact JSON (mcp_tools.tool_text, also for attached answers and run status), the canvas
leaves out the page's own "$$" widgets, and recent_activity gives the last 10 by default.
begin_task also tells the AI that AMB's tools cover models, pictures, nodes, canvas and runs,
so it need not read ComfyUI's folders, source or routes (each asks the user to allow it).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- Each run keeps its shape read from the prompt ComfyUI queued (ai_run_map.py): boxes by role
with their settings, typed links, plumbing wired through, and what changed since the task's
run before. The run step shows a one-line summary, the changes as chips and the boxes (top
down, fitted to the column, each locating its node); show_run_map puts it on the stage.
- By default the AI works in a tab of the task's own (ai_task_canvas.py, mcp_task_tab.js): the
open workflow is copied before its first change, the tab is brought back (a click on
ComfyUI's tab) before later changes, reads and runs, and the user's tab stays untouched.
Per-task switch in the header, default in the Gates window; the AI is told once.
- Output pictures can be handed to the task an AI is on (gallery card menu, the inspector's
✦), with the user's words: the AI gets them attached with what made them (ai_pictures.py).
search_images(with_pictures) attaches the first six.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- A stop button on a running run (timeline, and beside the run-status button), and stop_run for
the AI: only this run's prompts leave the queue or are interrupted; the step ends "cancelled"
and a user's stop reaches the AI as a "stop" message (ai_run_control.py).
- The run's progress carries ComfyUI's console since it began (the bar's note, models loading),
time left and per step, and VRAM; once queued, each sampler's working size is estimated from
the prompt, with warnings for an upscaled picture sampled again, a size past the card and a
huge decode, for the user and in the AI's run news (ai_run_watch.py, ui_ai_run_live.js).
- The materials picker opens as a side panel that leaves the page usable (the full-screen
browser makes room), expands to the large window and back, and remembers the choice.
- Esc closes the run-status card only (data-anomalous-escape).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The bar under a run shows a percent like ComfyUI's tab title (the running
node's steps, else the nodes done). While a run goes, every other tool answer
carries your_run with one sentence on where it stands, and the next answer
after it ends says so once (run_status telling it counts), so the AI need not
look in the output folder or at the screen.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- ai_task_runs.py reads where a run stands from ComfyUI's queue and progress registry (never
changing them): queued and how many are ahead, or the node running, its steps and the nodes
done. It is kept on the step in memory and sent as "progress" events that only move the
step's bar; the run's step is saved only when something new is kept.
- run_status (read-only): the task's latest run and its progress, its pictures once done;
wait_seconds (<= 55) waits for the end. run_workflow's answer tells the AI to use it instead
of guessing from the time.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- api/action_catalog.py: every MCP tool's kind, stage of the work and gate in one place (a test
keeps it and the tools in step); steps now carry their stage.
- api/ai_task_changes.py: changes the user makes in Anomalous while an AI works become steps of
its task, reach the AI once with its next call (kind "change": keep it), and the activity
entry names the task; the activity page links to it ("AI task ->").
- api/ai_task_runs.py: run watching moved out of ai_tasks.py (600-line rule); a step a restart
left "running" is marked stopped when the record is read again.
- Pictures: inspector / compare with canvas / keep as tools on hover instead of three buttons
under every picture; tick any two (any step or task) and compare from the bar at the foot of
the page; the results strip shows thumbnails again.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
