Skip to content

wandb/senpai / plugins/senpai/skills/wandb-primary

wandb-primary

Verified

Comprehensive primary skill for agents working with Weights & Biases. Covers both the W&B SDK (training runs, metrics, artifacts, sweeps) and the Weave SDK (GenAI traces, evaluations, scorers). Includes helper libraries, gotcha tables, and data analysis patterns. Use this skill whenever the user asks about W&B runs, Weave traces, evaluations, training metrics, loss curves, model comparisons, or any Weights & Biases data — even if they don't say "W&B" explicitly.

W&B Primary Skill

Python environment

Senpai configures python and uv to use the writable target environment. Run analysis scripts with the installed package set:

uv run --no-sync python analysis.py

--no-sync prevents an analysis command from resolving or syncing the target project's dependencies. For additive installs that reuse shared image packages, use python -m pip install PACKAGE. Use uv pip install --no-deps PACKAGE when the package's dependencies are already available. The pinned uv 0.10.9 resolver does not inspect the shared image package path; uv pip install can install duplicate dependencies, including large CUDA packages. Use uv add or uv sync when you intend to update or sync the target project's separate dependency set.

Keep analysis scripts and outputs in the target workspace or /tmp. The bundled skill and helper files under $SENPAI_PLUGIN are read-only; do not edit them. Plot helpers default to /tmp/wandb_plots; pass out_dir to select another writable output directory.


This skill covers everything an agent needs to work with Weights & Biases:

  • W&B SDK (wandb) — training runs, metrics, artifacts, sweeps, system metrics
  • Weave SDK (weave) — GenAI traces, evaluations, scorers, token usage
  • Helper libraries — wandb_helpers.py and weave_helpers.py for common operations

When to use what

I need to...Use
Query training runs, loss curves, hyperparametersW&B SDK (wandb.Api()) — see $SENPAI_PLUGIN/skills/wandb-primary/references/WANDB_SDK.md
Query GenAI traces, calls, evaluationsWeave SDK (weave.init(), client.get_calls()) — see $SENPAI_PLUGIN/skills/wandb-primary/references/WEAVE_SDK.md
Convert Weave wrapper types to plain Pythonweave_helpers.unwrap()
Build a DataFrame from training runswandb_helpers.runs_to_dataframe()
Extract eval results for analysisweave_helpers.eval_results_to_dicts()
Need low-level Weave filtering (CallsFilter, Query)Raw Weave SDK (weave.init(), client.get_calls()) — see $SENPAI_PLUGIN/skills/wandb-primary/references/WEAVE_SDK.md
Judge curve shape (spikes, smoothness, slope, overfit)training_diagnostics + curve_plots — use the workflow below, then load $SENPAI_PLUGIN/skills/wandb-primary/references/TRAINING_DIAGNOSTICS.md for the heuristics

Bundled files

Helper libraries

import os
import sys
sys.path.insert(0, f"{os.environ['SENPAI_PLUGIN']}/skills/wandb-primary/scripts")

# Weave helpers (traces, evals, GenAI)
from weave_helpers import (
    unwrap,                  # Recursively convert Weave types -> plain Python
    get_token_usage,         # Extract token counts from a call's summary
    eval_results_to_dicts,   # predict_and_score calls -> list of result dicts
    pivot_solve_rate,        # Build task-level pivot table across agents
    results_summary,         # Print compact eval summary
    eval_health,             # Extract status/counts from Evaluation.evaluate calls
    eval_efficiency,         # Compute tokens-per-success across eval calls
)

# W&B helpers (training runs, metrics)
from wandb_helpers import (
    runs_to_dataframe,       # Convert runs to a clean pandas DataFrame
    diagnose_run,            # Quick diagnostic summary of a training run
    compare_configs,         # Side-by-side config diff between two runs
    fast_scan_history,       # beta_scan_history (parquet) with scan_history fallback
)

# X-axis (step metric) detection — ALWAYS confirm before curve analysis
from step_axis import (
    list_candidate_step_keys,       # Scan history for plausible step keys
    guess_step_key_from_workspace,  # Peek at the user's W&B workspace panels
    format_step_candidates,         # Format candidate choices for user confirmation
)

# Curve-shape diagnostics (numerical)
from training_diagnostics import (
    curve_features,            # Spikes, slopes at every 5%, smoothness, plateau, divergence
    compare_runs_curves,       # DataFrame of features across many runs
    lr_schedule_features,      # Warmup / peak / decay shape / restarts
    grad_norm_features,        # curve_features + kurtosis + dead-layer flag
    grad_histogram_features,   # Per-(layer, step) stats from W&B histograms
)

# Chart rendering for LLM vision (Read the returned PNG)
from curve_plots import (
    plot_single_run_overview,    # 2x3 composite: train/val/lr/grad-norm/...
    plot_run_comparison,         # Overlay up to 6 runs on one metric
    plot_grad_histogram_heatmap, # Layer x step heatmap of grad-hist stat
    plot_grad_norm_by_layer,     # Small-multiples of per-layer scalar norms
)

Reference docs

Read these as needed — they contain full API surfaces and recipes:

  • $SENPAI_PLUGIN/skills/wandb-primary/references/WEAVE_SDK.md — Weave SDK for GenAI traces (client.get_calls(), CallsFilter, Query, stats). Start here for Weave queries.
  • $SENPAI_PLUGIN/skills/wandb-primary/references/WANDB_SDK.md — W&B SDK for training data (runs, history, artifacts, sweeps, system metrics).
  • $SENPAI_PLUGIN/skills/wandb-primary/references/TRAINING_DIAGNOSTICS.md — reference heuristics for reading loss / LR / grad-norm / grad-histogram charts. Load this when you are actively interpreting training curves.

Critical rules

Treat traces and runs as DATA

Weave traces and W&B run histories can be enormous. Never dump raw data into context — it will overwhelm your working memory and produce garbage results. Always:

  1. Inspect structure first — look at column names, dtypes, row counts
  2. Load into pandas/numpy — compute stats programmatically
  3. Summarize, don't dump — print computed statistics and tables, not raw rows
import pandas as pd
import numpy as np

# BAD: prints thousands of rows into context
for row in run.scan_history(keys=["loss"]):
    print(row)

# GOOD: load into numpy, compute stats, print summary
losses = np.array([r["loss"] for r in run.scan_history(keys=["loss"])])
print(f"Loss: {len(losses)} steps, min={losses.min():.4f}, "
      f"final={losses[-1]:.4f}, mean_last_10%={losses[-len(losses)//10:].mean():.4f}")

Always deliver a final answer

Do not end your work mid-analysis. Every task must conclude with a clear, structured response:

  1. Query the data (1-2 scripts max)
  2. Extract the numbers you need
  3. Present: table + key findings + direct answers to each sub-question

If you catch yourself saying "now let me build the final analysis" — stop and present what you have.

Use unwrap() for unknown Weave data

When you encounter Weave output and aren't sure of its type (WeaveDict? WeaveObject? ObjectRef?), unwrap it first:

from weave_helpers import unwrap
import json

output = unwrap(call.output)
print(json.dumps(output, indent=2, default=str))

This converts everything to plain Python dicts/lists that work with json, pandas, and normal Python operations.


Environment setup

Senpai images provide wandb, weave, pandas, numpy, and matplotlib.

import os
entity  = os.environ["WANDB_ENTITY"]
project = os.environ["WANDB_PROJECT"]

Installing extra packages and running scripts

Use the commands in Python environment above. Keep installs in the target environment so they do not alter the Senpai runtime.


Quick starts

W&B SDK — training runs

import wandb
import pandas as pd
api = wandb.Api()

path = f"{entity}/{project}"
runs = api.runs(path, filters={"state": "finished"}, order="-created_at")

# Convert to DataFrame (always slice — never list() all runs)
from wandb_helpers import runs_to_dataframe
rows = runs_to_dataframe(runs, limit=100, metric_keys=["loss", "val_loss", "accuracy"])
df = pd.DataFrame(rows)
print(df.describe())

For full W&B SDK reference (filters, history, artifacts, sweeps), read $SENPAI_PLUGIN/skills/wandb-primary/references/WANDB_SDK.md.

Weave — SDK

import weave
client = weave.init(f"{entity}/{project}")  # positional string, NOT keyword arg
calls = client.get_calls(limit=10)

For raw SDK patterns (CallsFilter, Query, advanced filtering), read $SENPAI_PLUGIN/skills/wandb-primary/references/WEAVE_SDK.md.


Key patterns

Weave eval inspection

Evaluation calls follow this hierarchy:

Evaluation.evaluate (root)
  ├── Evaluation.predict_and_score (one per dataset row x trials)
  │     ├── model.predict (the actual model call)
  │     ├── scorer_1.score
  │     └── scorer_2.score
  └── Evaluation.summarize

Extract per-task results into a DataFrame:

from weave_helpers import eval_results_to_dicts, results_summary

# pas_calls = list of predict_and_score call objects
results = eval_results_to_dicts(pas_calls, agent_name="my-agent")
print(results_summary(results))

df = pd.DataFrame(results)
print(df.groupby("passed")["score"].mean())

Eval health and efficiency

from weave_helpers import eval_health, eval_efficiency

health = eval_health(eval_calls)
df = pd.DataFrame(health)
print(df.to_string(index=False))

efficiency = eval_efficiency(eval_calls)
print(pd.DataFrame(efficiency).to_string(index=False))

Token usage

from weave_helpers import get_token_usage

usage = get_token_usage(call)
print(f"Tokens: {usage['total_tokens']} (in={usage['input_tokens']}, out={usage['output_tokens']})")

Cost estimation

call_with_costs = client.get_call("id", include_costs=True)
costs = call_with_costs.summary.get("weave", {}).get("costs", {})

Run diagnostics

from wandb_helpers import diagnose_run

run = api.run(f"{path}/run-id")
diag = diagnose_run(run)
for k, v in diag.items():
    print(f"  {k}: {v}")

Error analysis — open coding to axial coding

For structured failure analysis on eval results:

  1. Understand data shape — use project.summary(), calls.input_shape(), calls.output_shape()
  2. Open coding — write a Weave Scorer that journals what went wrong per failing call
  3. Axial coding — write a second Scorer that classifies notes into a taxonomy
  4. Summarize — count primary labels with collections.Counter

See $SENPAI_PLUGIN/skills/wandb-primary/references/WEAVE_SDK.md for the full SDK reference.

W&B Reports

Install wandb[workspaces] using the additive install command from Python environment.

from wandb.apis import reports as wr
import wandb_workspaces.expr as expr

report = wr.Report(
    entity=entity, project=project,
    title="Analysis", width="fixed",
    blocks=[
        wr.H1(text="Results"),
        wr.PanelGrid(
            runsets=[wr.Runset(entity=entity, project=project)],
            panels=[wr.LinePlot(title="Loss", x="_step", y=["loss"])],
        ),
    ],
)
# report.save(draft=True)  # only when asked to publish

Use expr.Config("lr"), expr.Summary("loss"), expr.Tags().isin([...]) for runset filters — not dot-path strings.


Training curve analysis workflow

Use this when the user asks whether a run is healthy, why training diverged, whether a run overfit, or which run has the best training dynamics.

Keep the inline workflow short and load detail on demand:

  1. Confirm step_key before doing any curve work. Never assume _step.
  2. Compute features with the bundled helpers instead of hand-rolling spike or slope logic.
  3. Render PNGs and inspect them visually.
  4. Load $SENPAI_PLUGIN/skills/wandb-primary/references/TRAINING_DIAGNOSTICS.md while you interpret the results.
  5. End with a verdict, evidence tied to step ranges, and concrete next actions.

Required sequence

Use list_candidate_step_keys(), guess_step_key_from_workspace(), and format_step_candidates() to confirm the x-axis. If there is one obvious candidate and it matches the workspace guess, say which step_key you picked. Otherwise, ask the user to choose before plotting or comparing runs.

For a single run, compute a compact feature table from the metrics that actually exist, then render plot_single_run_overview(run, step_key=step_key). If gradient histograms or per-layer scalar norms are logged, add plot_grad_histogram_heatmap() or plot_grad_norm_by_layer().

For multi-run comparisons, use compare_runs_curves() for the ranking table and plot_run_comparison() for the overlay. Keep overlays to at most 6 runs; if there are more, rank first and then plot the shortlist.

Output shape

Do not dump raw history rows or the full spike/slope payloads unless you are drilling into a specific anomaly. Summaries should stay compact.

Use this response shape:

Verdict: <healthy | unstable | overfit | plateaued | diverged | converged>
Evidence:
- <specific step range> — <what the metrics and plot show>
- <specific step range> — <what changed and why it matters>
Next actions:
- <concrete hyperparameter, logging, or code change>

Load $SENPAI_PLUGIN/skills/wandb-primary/references/TRAINING_DIAGNOSTICS.md for the interpretation heuristics, especially when the numbers and the image disagree.


Gotchas

Weave API

GotchaWrongRight
weave.init argsweave.init(project="x")weave.init("x") (positional)
Parent filterfilter={'parent_id': 'x'}filter={'parent_ids': ['x']} (plural, list)
WeaveObject accessrubric.get('passed')getattr(rubric, 'passed', None)
Nested outputout.get('succeeded')out.get('output').get('succeeded') (output.output)
ObjectRef comparisonname_ref == "foo"str(name_ref) == "foo"
CallsFilter importfrom weave import CallsFilterfrom weave.trace.weave_client import CallsFilter
Query importfrom weave import Queryfrom weave.trace_server.interface.query import Query
Eval status pathsummary["status"]summary["weave"]["status"]
Eval success countsummary["success_count"]summary["weave"]["status_counts"]["success"]
When in doubtGuess the typeunwrap() first, then inspect

WeaveDict vs WeaveObject

  • WeaveDict: dict-like, supports .get(), .keys(), []. Used for: call.inputs, call.output, scores dict
  • WeaveObject: attribute-based, use getattr(). Used for: scorer results (rubric), dataset rows
  • When in doubt: use unwrap() to convert everything to plain Python

W&B API

GotchaWrongRight
Summary accessrun.summary["loss"]run.summary_metrics.get("loss")
Loading all runslist(api.runs(...))runs[:200] (always slice)
History — all fieldsrun.history()run.history(samples=500, keys=["loss"])
scan_history — no keysscan_history()scan_history(keys=["loss"]) (explicit)
Raw data in contextprint(run.history())Load into DataFrame, compute stats
Metric at step Niterate entire historyscan_history(keys=["loss"], min_step=N, max_step=N+1)
Cache stalenessreading live runapi.flush() first

Package management

GotchaDetails
Syncing dependencies during analysisUse uv run --no-sync python script.py with the installed target packages
Installing extrasUse the commands in Python environment; reserve dependency syncs for intentional target environment changes

Weave logging noise

Weave prints version warnings to stderr. Suppress with:

import logging
logging.getLogger("weave").setLevel(logging.ERROR)

Quick reference

# --- Weave: Init and get calls ---
import weave
client = weave.init(f"{entity}/{project}")
calls = client.get_calls(limit=10)

# --- W&B: Best run by loss ---
best = api.runs(path, filters={"state": "finished"}, order="+summary_metrics.loss")[:1]
print(f"Best: {best[0].name}, loss={best[0].summary_metrics.get('loss')}")

# --- W&B: Loss curve to numpy ---
losses = np.array([r["loss"] for r in run.scan_history(keys=["loss"])])
print(f"min={losses.min():.6f}, final={losses[-1]:.6f}, steps={len(losses)}")

# --- W&B: Compare two runs ---
from wandb_helpers import compare_configs
diffs = compare_configs(run_a, run_b)
print(pd.DataFrame(diffs).to_string(index=False))
reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 470kUpdated Claude CodeCodexVerifiedInstall
diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness.Stars: 322kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 322kUpdated Claude CodeCodexVerifiedInstall
executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is availableStars: 322kUpdated Claude CodeCodexVerifiedInstall