Est. Cost
$0.00
Actual cost may be lower with prompt caching
Active Sessions
0
Total Calls
0
Cache Savings
$0.00
Cloud Calls
0
Local Calls
0
PII Detections
0
Guardrail Detections
0
Cache Hits
0

Live Event Stream

No events yet. Make an LLM call to see activity here.

Sessions

Session ID Status Est. Cost Tokens Calls Cache Hits PII Org Team Started
Page 1

Async Jobs

Total Jobs
0
Running
0
Completed
0
Failed
0
Job ID Status Provider Model Team Session Created Started Completed Retries Error

Models

Cloud

Emergency Model Override
Saved!
Total Cloud Spend
$0.00
Cloud Calls
0
Top Model
—
Cost by Model

Provider API Keys

OpenAI
Not configured
Anthropic
Not configured
Google (Gemini)
Not configured

Local

Ollama Status
Offline
Force Local Routing
Auto-Route
Active Local Model

Downloaded Models

Model Size Modified Actions

Hot-Swap Model

New Model
Delete Old Model

Recommended Models

Test candidate models against your production cloud model automatically. When enabled, every cloud LLM call also fires a parallel call to the candidate model in the background — no explicit traffic needed. Just use your app normally and StateLoom compares responses for you.

Configuration

Model Testing
Candidate Model
Sampling Rate
100%
Max Context Tokens
Similarity Method

Test Metrics

Test Runs
0
Success Rate
0%
Cost Saved
$0.00
Avg Similarity
—

Filtering Funnel

Total Calls 0
Eligible 0
Tested 0

Readiness Report

Run model tests to generate readiness scores. Enable testing above, select a candidate model, then use your app normally.

Enterprise Roadmap

Enterprise
🔒
Dark Launching

Mirror live production traffic to candidate models. Check functional equivalence — not just text similarity: valid JSON schemas, correct tool calls, matching response structure. Get a live migration confidence score.

Enterprise
🔒
Distillation Flywheel

Automatically generate fine-tuning datasets from your production traffic. When a local SLM fails where the cloud model succeeds, StateLoom collects the prompt + correct response as a Chosen/Rejected training pair in .jsonl format.

Session Detail:

Status
-
Total Cost
$0.00
Total Tokens
0
Prompt Tokens
0
Completion Tokens
0
Calls
0
Cache Hits
0
Cache Savings
$0.00
PII Detections
0
Guardrail Detections
0
Retries
0
Organization
—
Team
—

Trace Timeline

Step Type Model Prompt Duration Cost Status

Experiments

ID Name Status Strategy Variants Assignments Created
Rank Experiment Status Variant Success Rate Avg Cost

Agents

Total
0
Active
0
Paused
0
Archived
0
Slug Name Status Model Team Version Created

Create New Agent

Team
Slug a-z, 0-9, hyphens, 3-64 chars
Name
Description
Model
System Prompt
Request Overrides
Budget / Session
← Back to Agents

Proxy URL

Active Version

Version History

Version Model Created Created By Actions

Virtual Keys

PreviewNameCreatedActions

Consensus Runs

Total Runs
0
Avg Confidence
-
Est. Cost
$0
Early Stopped
0
Session Strategy Models Rounds Confidence Cost Duration Time
Page 1

Compliance

PII Monitor

Total Detections
0
Sessions Affected
0
Types Detected
—
Actions Taken
—
Timestamp Session Type Detected Value Action Field
Page 1

Enterprise PII Rules

PII Enabled
Default Mode
Active Rules
0
Pattern Mode On Failure Actions
Loading profiles...
Timestamp Standard Action Legal Rule Target Org ID Integrity Hash

Purge User Data

User Identifier
Compliance Standard
Organizations
0
Teams
0
Loading organizations...

Security

Global Kill Switch
Response Mode
Active Rules
0
Environment
—

Kill Switch Message

Updated!

Rules

Model Provider Environment Agent Version Message Reason
Enabled
—
Paused Sessions
0
Paused Agents
0

Paused Sessions

No paused sessions

Paused Agents

No paused agents

Failure Counts

Session Failures
—
Agent Failures
—

Recent Events

No blast radius events
Guardrails
Disabled
Mode
—
Total Detections
0
Blocked
0
Heuristic Patterns
Disabled
NLI Classifier
Disabled
Local Model
Disabled
Output Scanning
Disabled
Pattern Count
0

Detections by Category

Detections by Severity

Recent Violations

Timestamp Rule Category Severity Phase Validator Action
Audit Hooks
Mode
Events Logged
0
Blocked
0

Deny Events

Operations that trigger audit/block when detected.

Allow Paths

Glob patterns for file paths that bypass file open checks.

Vault Status
Disabled
Vaulted Keys
0
Scrubbed from Env
0

Vaulted Keys

Key Name Status

Add Secret

Timestamp Event Detail Action

Observability

Auto-refresh: 30s
p50 Latency
—
p90 Latency
—
p95 Latency
—
p99 Latency
—

Request Rate

Latency Distribution

Cost Over Time

Token Usage

Cache Performance

Provider Breakdown

Top Models

Settings

Budget

Per-Session Budget
Global Budget
Budget Action
Saved!

Cache

Enabled restart required -
Max Size
Saved!

Loop Detection

Enabled
Exact Threshold
Saved!

Storage

Retention Days
Saved!

Console

Output restart required -
Verbose
Saved!

Server Logs

No logs captured yet.