argus-agents 0.9.1

Catch silent failures in AI agent pipelines before production.

Your LangGraph pipeline runs fine — no exception. But three nodes later, something crashes with a KeyError. The real cause? A node upstream silently dropped a field. ARGUS catches this.

Install

pip install argus-agents

This is the full product: the argus CLI, the LangGraph adapter, and the local UI (argus ui). ARGUS runs fully local — runs are stored in .argus/runs/, no account, no cloud, no signup. Heuristic detection (150+ signatures) works out of the box.

The PyPI package is argus-agents, not argus.

Bring Your Own Key (BYOK)

AI-powered detection (the semantic judge, LLM investigator, learned trends) uses your own key from the provider of your choice — OpenAI, Anthropic (Claude), or Google (Gemini). Set it once and it's saved locally for every future session:

argus key set                          # OpenAI by default — prompts, hidden input
argus key set --provider anthropic     # or Anthropic (Claude)
argus key set --provider google        # or Google (Gemini)
# pass it directly instead of being prompted:
argus key set sk-... --provider openai
# or just export it (env wins over the saved key):
export OPENAI_API_KEY=sk-...           # or ANTHROPIC_API_KEY / GEMINI_API_KEY

Configured more than one? Switch the active provider anytime:

argus key use anthropic  # activate a provider you already have a key for
argus key show           # list configured providers (masked); * marks the active one
argus doctor             # reports BYOK provider / hosted / heuristic-only mode

You pick the provider; ARGUS picks a sensible balanced model for each internal call (a cheap model for the frequent per-node checks, a stronger one for root-cause reasoning). Per-provider resolution order: env var (OPENAI_API_KEY / ANTHROPIC_API_KEY / GEMINI_API_KEY) → saved key → (hosted proxy, if you're on the cloud tier) → heuristic-only.

No key? ARGUS still works — it falls back to heuristic-only detection, no crashes.

Getting Started (AI-Powered Setup)

The fastest way to set up ARGUS — let your AI assistant handle the integration:

1. Go to arguslabs.in

2. Click on "AI Setup Prompt"

3. Copy the prompt and paste it into your AI assistant (Claude, ChatGPT, Cursor, etc.) — it will handle the full setup for you. This is the most important step.

4. (Optional) Log in for hosted cloud sync

argus login

Signing in with Google enables hosted cloud sync and the shared trends registry — part of the managed/enterprise tier. It is entirely optional: the open-source package is fully local and needs no login. (argus login reports "hosted-only feature" unless a hosted backend is configured.)

5. Open the dashboard

argus ui

That's it. ARGUS is set up and ready to go.

After Setup

• Run your LangChain/LangGraph pipeline as usual (in terminal)
• Check the dashboard for your recent run (takes 1–2 seconds to appear — refresh if needed)

Quick Start

from argus import ArgusWatcher

watcher = ArgusWatcher()
app = watcher.attach(graph)         # StateGraph or already-compiled app
result = app.invoke(initial_state)  # run is persisted automatically

ARGUS monitors every node, detects failures, and saves the run. No changes to your node functions.

finalize() is optional. attach() wraps invoke() / ainvoke() / batch() / abatch() / stream() so the run is written to .argus/runs/ when the outermost call returns — including cyclic graphs. Calling watcher.finalize() afterwards is a no-op.

Constructor form still works if you compile yourself:

watcher = ArgusWatcher(graph)       # uncompiled StateGraph
app = graph.compile()
result = app.invoke(initial_state)

What It Catches

Detection Layers

Runs in order, each more expensive — only fires when needed:

1. Heuristics — 150+ failure signatures (placeholders, empty results, error keys, semantic degradation). Zero cost.
2. Validators — custom per-node business-logic constraints. Deterministic.
3. Anomaly detector — statistical checks for output size anomalies, timing outliers. Deterministic.
4. Correlator — traces failure propagation across nodes. Points at the origin, not the crash site.
5. LLM semantic judge — evidence-aware final ruling. Receives all signals from layers 1–4 before deciding. Cannot override validator failures or critical anomalies.
6. LLM investigator — root cause explanations and debugging suggestions. Only on ambiguous failures.
7. Loop analyzer — LLM analysis for looped nodes: summarizes iterations, detects stalls, flags wasted retries.

Loop-Aware Inspection

Pipelines with loops (LLM -> compiler -> if fail, retry) get special treatment:

• Earlier iterations that self-corrected are marked retried (not counted as failures)
• Only the final iteration determines pass/fail
• LLM analyzes every loop: what went wrong, what changed between attempts, whether retries were necessary
• Dashboard shows iteration badges, collapse/expand, and natural-language loop summaries

Replay

Fix a bug, re-run from the failing node. Skip upstream nodes entirely:

argus replay <run-id> node_7          # re-run from node_7 onward
argus replay <run-id> node_7 --only   # just that one node
argus diff <rerun-id>                 # compare vs original

External API calls (OpenAI, etc.) are recorded by default — replays are free and deterministic.

Semantic Judge

For subtle quality issues that pattern matching can't catch:

watcher = ArgusWatcher(graph, semantic_judge=True)  # enabled by default

LLM evaluates output quality on every node. Catches wrong tone, unhelpful responses, outdated info. Requires a provider key (OpenAI, Anthropic, or Google) — set via argus key set [--provider ...] (see BYOK).

The judge receives all prior evidence — validator failures, anomaly signals, inspection results — so it rules with full context, not just input/output. Every decision includes an audit trail:

{
  "pass": false,
  "reason": "Validator correctly identified missing resolution_ticket",
  "confidence": 0.85,
  "evidence_considered": ["validator:payment_check", "anomaly:BA-003"],
  "overridden_signals": []
}

• evidence_considered — which prior signals the LLM weighed
• overridden_signals — which signals the LLM disagreed with (passed despite the flag)

Custom Validators

watcher = ArgusWatcher(graph, validators={
    "classify": lambda o: (o.get("label") in ["yes", "no"], "unexpected label"),
    "*":        lambda o: ("error" not in o, "error key present"),  # runs on every node
})

Validator failures cannot be overridden by the LLM judge — they are hard constraints.

Configuration

from argus import ArgusWatcher, ArgusConfig

config = ArgusConfig(
    semantic_judge=True,           # LLM judge on every node (default: True)
    judge_model="gpt-4o",          # model for the judge
    node_timeout_ms=30000,         # flag outputs at ≥95% of this
    min_expected_ms=500,           # flag suspiciously fast LLM nodes
    sample_rate=0.5,               # persist 50% of clean runs (save disk)
    persist_failures=True,         # always persist failed runs
)

watcher = ArgusWatcher(graph, config=config)

CLI

argus list                           # all recorded runs
argus show last                      # most recent run
argus show <id>                      # inspect a specific run
argus inspect <id> --step <node>     # dump raw input/output for a node
argus replay <id> <node>             # re-run from a node
argus diff <id-a> <id-b>             # compare two runs
argus stats                          # signature hit stats, disable/enable/dispute signatures
argus ui                             # web dashboard
argus doctor                         # check setup health + LLM mode (BYOK/hosted/heuristic)
argus key set [--provider ...]       # save a provider key locally (OpenAI/Anthropic/Google) — BYOK
argus key use <provider>             # switch the active provider
argus key show                       # list configured providers (masked); * marks active
argus key clear [--provider ...]     # remove one provider's key, or all
argus login                          # (optional) sign in for hosted cloud sync
argus logout                         # clear stored credentials
argus whoami                         # show current login status
argus update                         # check for newer release

Web Dashboard

argus ui    # opens at localhost:7842

Shows all runs, node-level detail, AI analysis, replay diffs, loop iteration badges, and comparison views. No account needed for local use.

• Distinct failure colors — crashed (red), silent failure (amber), semantic fail (purple), degraded input (orange), skipped (gray)
• Evidence audit trail — see exactly which signals the LLM judge considered and which it overrode
• Side-by-side diff — compare any two runs node-by-node

Without LangGraph

from argus import ArgusSession

session = ArgusSession()
session.set_edges({"fetch": ["classify"], "classify": ["process"]})

fetch    = session.wrap("fetch",    fetch_fn)
classify = session.wrap("classify", classify_fn)
process  = session.wrap("process",  process_fn)

state = fetch(initial_state)
state = classify(state)
state = process(state)
session.finalize()

Works with any framework — Prefect, Temporal, plain Python.

Requirements

• Python 3.9+
• LangGraph 0.2+ (only for ArgusWatcher)
• A provider key (OpenAI, Anthropic, or Google) for semantic features — set via argus key set [--provider ...] (optional; all heuristic detection works without it)

For AI setup prompts and integration guides, visit arguslabs.in.

v0.8.12 — changelog

License

ARGUS is open-core. The open-source core (src/argus/, the argus-agents PyPI package) is licensed under Apache-2.0 — see LICENSE. The cloud/ directory (hosted/enterprise components) is proprietary — see cloud/LICENSE.