Workspace: northstar · TEAM plan · 10,000,000 events/month
Ask spend, AI contribution margin, waste, recommendation, budget, and alert questions from Slack, Cursor, another MCP client, or the trAIce CLI.
trAIce posts to Slack through a Slack Incoming Webhook URL.
Treat the webhook URL like a secret. Slack ties each URL to one workspace and channel.
| Alert | Severity | Enabled | Cadence | Channels | Thresholds |
|---|---|---|---|---|---|
Monthly usage nearing quota System alert | Warning | ||||
Monthly usage quota reached System alert | Critical | ||||
Workspace seat added System alert | Digest | ||||
Workspace plan changed System alert | Digest | ||||
Workspace role changed System alert | Digest | ||||
Workspace member removed System alert | Digest | ||||
API key changed System alert | Warning | ||||
Alert delivery settings changed System alert | Warning | ||||
Scheduled reduction needs action System alert | Warning | ||||
Scheduled reduction cancelled System alert | Warning | ||||
Model rollout automatically rolled back Rollout alert | Critical | ||||
Customer flipped to negative margin Margin alert | Critical | ||||
Customer margin below floor Margin alert | Warning | ||||
Blended margin below target Margin alert | Warning | ||||
New #1 margin leak Margin alert | Digest | ||||
Attributed spend spike Cost alert | Warning | ||||
Budget threshold + projected overage Cost alert | Warning | ||||
Retry / error waste spike Waste alert | Warning | ||||
Runaway run Waste alert | Warning | ||||
Run stuck in a loop Waste alert | Warning | ||||
New validated savings available Savings alert | Digest | ||||
Applied swap silently reverted Savings alert | Warning | ||||
Ingestion stopped / dropped Data alert | Critical | ||||
Untagged spend rising Data alert | Warning | ||||
Revenue mapping missing or stale Data alert | Warning |
Demo path: open the margin leak, then inspect chat-resolver logs for Initech. In production this is where you would reprice, throttle, or route to a cheaper model.
Product AI monthly cap is at 30.2% of cap; on pace for $56,083 vs $55,000. Enforcement: advisory policy available in SDK config.
Product AI spend is at 91.1% of its monthly cap and projected to exceed it. Review the budget before the next support peak.
Chat resolver spend rose faster than traffic. Review the 11-day-ago spike and the validated route before changing the budget.
Liam O'Connor crossed Helpdesk Core · $496 per employee. trAIce detected the employee-level overage inside Helpdesk Core; review the spike before it becomes the team baseline.
The 30-case ticket-summary experiment passed on Claude Haiku 4.5. Review the measured cases and continue the progressive rollout.
Owners and admins can allow trAIce to retain redacted text prompt and output pairs. Each captured pair expires up to 14 days after it is created. Starting a run copies 30 selected pairs into encrypted evidence with a separate 14-day expiry. Eligible samples are sent to Google Vertex AI and the selected candidate provider. System instructions, tools, retrieval context, attachments, and agent state are not captured and are not supported by these experiments. Opt out to delete retained samples and evidence immediately.
Off by default. When enabled, each experiment sends the same 30 frozen prompts to the original source provider again so source and candidate latency can be compared from the same worker. This can double model generation calls and adds source-provider spend. The new source output is discarded after timing and usage are recorded. The captured production response remains the quality reference. Opting out stops future source legs in an active run after any in-flight provider request finishes.
By default, experiments compare parsed JSON content and report presentation adherence separately. Add an exact feature name only when whitespace, code fences, or another presentation detail is part of the real downstream contract. Changes apply to future experiment runs.
SDK must also be configured to send prompt/output on event payloads. See docs.
Shadow and recommendation rules only record would-be actions. Active wrapper-v1 rules can run through @traice/sdk. Swap and downgrade rules require current experiment evidence.
| Name | When | Then | State | Actions |
|---|---|---|---|---|
| Fallback when the primary provider is unavailable | feature = ticket-summary | fallback → gpt-4o | active · fallback | |
| Block requests beyond tenant policy | feature = chat-resolver | deny | active · deny | |
| Cap repeated knowledge retrieval | feature = knowledge-agent | cap retries → gpt-5.5 | active · cap retries | |
| Reuse semantically equivalent knowledge answers | feature = knowledge-agent | cache semantic | active · cache semantic | |
| Reuse exact helpdesk responses | feature = chat-resolver | cache exact | active · cache exact | |
| Route low-risk chat requests | feature = chat-resolver | route → gpt-4o (≥90% eq) | active · route | |
| Downgrade routine quality reviews | feature = quality-review | downgrade → gpt-4o (≥90% eq) | active · downgrade | |
| Use Haiku for validated ticket summaries | feature = ticket-summary | swap → claude-haiku-4-5 (≥90% eq) | active · swap |
Bind this workspace to one public repository and one side of the experiment. Run the exact prompt suite on both sides, with the same agent and model settings.
Public View mode creates a read-only, unlisted link in the standard app shell. It shows the repository, deliberately configured prompt suite, experiment markers, token totals, anonymized agent labels, and timing. It does not expose captured prompts or outputs, local paths, people, teams, sources, credentials, raw model identifiers, raw metadata, or authenticated workspace access.
Rename this workspace without changing its slug, API keys, or integrations.
Download an NDJSON export of your workspaces before making lifecycle changes. Sensitive API key hashes are excluded.
Download export| Usage records | 161,680 |
| Imports | 1 |
| Workspace access | 132 |
| Active credentials | 130 |
| Configurations | 46 |
This is separate from workspace deletion. It immediately deletes your trAIce login, removes you from shared workspaces, and permanently deletes solo workspaces you own. Transfer sole ownership first. This action has no recovery period.
Type DELETE MY ACCOUNT to confirm:
| When | Rule | Action | Requested → Served | Mode | Saved |
|---|---|---|---|---|---|
| Aug 10, 04:33 AM | Use Haiku for validated ticket summaries | swap | claude-sonnet-4-6 → claude-haiku-4-5 | active | $12 |
| Aug 10, 03:56 AM | Use Haiku for validated ticket summaries | swap | claude-sonnet-4-6 → claude-haiku-4-5 | active | $12 |
| Aug 10, 03:19 AM | Use Haiku for validated ticket summaries | swap | claude-sonnet-4-6 → claude-haiku-4-5 | active | $12 |
| Aug 10, 02:42 AM | Use Haiku for validated ticket summaries | swap | claude-sonnet-4-6 → claude-haiku-4-5 | active | $12 |
| Aug 10, 02:40 AM | Downgrade routine quality reviews | downgrade | gpt-5.5 → gpt-4o | active | $10 |
| Aug 10, 02:05 AM | Use Haiku for validated ticket summaries | swap | claude-sonnet-4-6 → claude-haiku-4-5 | active | $12 |
| Aug 10, 02:03 AM | Downgrade routine quality reviews | downgrade | gpt-5.5 → gpt-4o | active | $10 |
| Aug 10, 01:28 AM | Use Haiku for validated ticket summaries | swap | claude-sonnet-4-6 → claude-haiku-4-5 | active | $12 |
| Aug 10, 01:26 AM | Downgrade routine quality reviews | downgrade | gpt-5.5 → gpt-4o | active | $10 |
| Aug 10, 12:51 AM | Use Haiku for validated ticket summaries | swap | claude-sonnet-4-6 → claude-haiku-4-5 | active | $12 |
| Aug 10, 12:49 AM | Downgrade routine quality reviews | downgrade | gpt-5.5 → gpt-4o | active | $10 |
| Aug 10, 12:47 AM | Route low-risk chat requests | route | gpt-5.5 → gpt-4o | active | $8 |
| Aug 10, 12:14 AM | Use Haiku for validated ticket summaries | swap | claude-sonnet-4-6 → claude-haiku-4-5 | active | $12 |
| Aug 10, 12:12 AM | Downgrade routine quality reviews | downgrade | gpt-5.5 → gpt-4o | active | $10 |
| Aug 10, 12:10 AM | Route low-risk chat requests | route | gpt-5.5 → gpt-4o | active | $8 |
| Aug 9, 11:53 PM | Use Haiku for validated ticket summaries | swap | claude-sonnet-4-6 → claude-sonnet-4-6 | active | - |
| Aug 9, 11:53 PM | Use Haiku for validated ticket summaries | swap | claude-sonnet-4-6 → - | active | - |
| Aug 9, 11:45 PM | Use Haiku for validated ticket summaries | swap | claude-sonnet-4-6 → claude-sonnet-4-6 | active | - |
| Aug 9, 11:45 PM | Use Haiku for validated ticket summaries | swap | claude-sonnet-4-6 → - | active | - |
| Aug 9, 11:37 PM | Use Haiku for validated ticket summaries | swap | claude-sonnet-4-6 → claude-sonnet-4-6 | active | - |
| Aug 9, 11:37 PM | Use Haiku for validated ticket summaries | swap | claude-sonnet-4-6 → - | active | - |
| Aug 9, 11:37 PM | Use Haiku for validated ticket summaries | swap | claude-sonnet-4-6 → claude-haiku-4-5 | active | $12 |
| Aug 9, 11:35 PM | Downgrade routine quality reviews | downgrade | gpt-5.5 → gpt-4o | active | $10 |
| Aug 9, 11:33 PM | Route low-risk chat requests | route | gpt-5.5 → gpt-4o | active | $8 |
| Aug 9, 11:29 PM | Use Haiku for validated ticket summaries | swap | claude-sonnet-4-6 → claude-sonnet-4-6 | active | - |