Z

Context Optimization Dashboard

Enter your API key to view compression analytics and performance metrics.

Current Showcase Run
Started: —
Run Requests
—
Run Tokens Processed
—
Run Tokens Saved
—
Run Savings Rate
—
Run Est. Savings
—
Current-run metrics are calculated from live account totals minus the baseline captured when this showcase started. Numbers are real measured deltas — never fake or hardcoded. Selected-period totals remain available in the cards below.
Enterprise flow mode: live requests are streaming; current-run totals update as traffic arrives.
Current-run baseline unavailable — showing selected-period totals.
Operator note: backend reported run tokens saved > run tokens processed. Run Savings Rate clamped to 100% pending stats-route investigation.
Last valid update: —
Waiting for latest stats; preserving last successful current-run values.
Current Run Model Mix
Live model breakdown for this run only.
—
Current Run Route Mix
Live route family breakdown for this run only.
—
Current Run Savings Trend
Tokens saved over the current run. One point per successful stats refresh — never interpolated.
no points yet
Projection at scale (projection, not actual)
Applies the current run's measured savings rate to enterprise token volumes. Not a guarantee. Independent of the actual Run Est. Savings card above.
—
Prompt families currently exercising Zur-lix optimization paths
Case-family activity for this run. Reflects which prompt shapes the runner is sending; not persisted layer-level telemetry.
—
Current Run Recent Requests
Backend recent_activity filtered to requests created on or after this run's baseline timestamp.
—
Z
Zur-lix
AI Efficiency Layer
Overview Analytics Requests Models Savings Savings View Settings
Account
—

Overview

Token optimization across models, requests, and savings — for the selected billing period.
Enterprise Flow Demo — showing this run only
Holding latest successful demo values
API Status Z
Metrics below are scoped to the selected billing period.

Rollout preparation status

Zur-lix is in controlled rollout preparation, not a public or global launch.
Savings figures
Informational estimates only — corpus-measured (typically ~16%, up to ~46–62% on favorable prompts), not guaranteed, and not a billing-grade figure unless promoted after go-live.
Billing
Billing is not active. No invoice and no charge is authorized. Billing readiness is operator-controlled.
API
Core API readiness is improving; streaming is not customer-contract-pinned yet.
Access
New external customers are not yet enabled; final activation is controlled by an explicit go-live decision.
Savings on your own traffic will vary and are not guaranteed. No billing action is authorized by this view.
Total Requests
—
Tokens Processed
—
Tokens Saved
—
Savings Rate
—
of input tokens recovered
Avg Latency
—
Est. Cost Savings
—
Total Customer Savings
—
Compression + Cache + Routing
Routing estimated — • actual: not yet available

Daily Token Savings

Requests by Model

OpenAI Anthropic Other

Model Breakdown

Route Breakdown

Direct Savings Sources

Decomposition of headline tokens_saved. Not added on top of headline savings.
Awaiting savings data…

Routing Decisions

Per-tenant routing-decision summary. Routing is fail-open; see safety note below.
Total Decisions
—
Swaps
—
No-Swap
—
Swap Rate
—
Total Saved
—
Avg / Swap
—
Top model pairs (swaps)
No swaps in this window.
Recent swaps
No recent swaps.
Routing summary is temporarily unavailable for this window. Last successful values shown when available; otherwise zero placeholders.

Account Status

Read-only account/subscription/discount/billing-period summary. Runs no billing, no Stripe calls, no provider calls, and authorizes nothing.
loading…
Use the current live billing period when verifying new demo traffic. Future periods may correctly show no usage yet.
Choose a billing period (YYYY-MM) and click Load account status. This card reflects the selected period only. It does not create usage, generate billing, push savings, or call Stripe.
Selected period — status —
Account active
—
Plan
—
API access
—
Subscription
—
Platform fee
—
Discount
—
Billing period summary
Period status
—
Billing-grade savings
—
Pending savings
—
Pushed savings
—
These figures are informational only — not a savings guarantee, not a billing-grade claim, and not an invoice or charge. "Billing-grade savings" is the reconciled-status field name, not a promise of a charge.
Next step
—
Read-only view of the selected billing period. No billing action is authorized by this card.

Observed Savings

Informational estimate from Zur-lix optimization layers. Not billing-grade. Not invoiced. No customer charge. Stripe / billing push: off. Raw prompt capture remains off. Realized savings vary by workload, prompt shape, model, and usage pattern. Authorizes nothing.
loading…
Click Load observed savings to view the latest informational estimate from the Zur-lix optimization layers. These numbers are informational only, not billing-grade, and do not authorize any customer charge or Stripe push. Waiting for measured optimization activity will show no current measured savings rather than fake totals.
Status
—
Rows observed
—
Observed tokens saved (informational total)
—
Average tokens saved / request
—
Observed savings are attributable to Zur-lix optimization layers (Context Relevance Optimization, Redundancy & Deduplication, Cache-Aware Context Shaping). They are informational only — not billing-grade unless explicitly promoted, not invoiced, not charged, and not pushed to Stripe. Actual savings vary by workload.

Billing Operator Tools

CLI remains source-of-truth; dashboard is visibility only.
Operator admin token
Used only in memory for read-only billing operator cards. Not stored. The token is never written to localStorage, sessionStorage, cookies, or the URL, and is never logged or rendered into result text.
Admin token not set

Savings Event Generation Preview pre-generation preview

Read-only. Does not call Stripe. Does not mutate DB. Preview only. Does not insert BillingSavingsEvent rows. Not an execute or generation surface. CLI (python -m app.operator.savings_generation_preview_cli) remains source-of-truth; dashboard is visibility only. Requires the shared Operator admin token above.
idle
Status
—
Usage events
—
Candidate usage events
—
Existing BSE rows
—
Would create
—
Would skip (existing)
—
Would create billing-grade
—
Would create micros
—
Would skip (non-positive)
—
Would skip (not billing-grade)
—
Read-only
—
Stripe call attempted
—
Rows inserted
—
Rows updated
—
All preview candidates already exist. Nothing new would be created.
No usage source rows found for this customer and period.
Preview indicates rows would be created by a later approved generation step. This card does not create them.
By axis
By price basis
By savings kind
Errors
Warnings
Click Run generation preview to fetch live state.

Savings Billing Reconciliation completed periods

Read-only. Does not call Stripe. Does not mutate DB. CLI (python -m app.operator.savings_reconciliation_readout_cli) remains the source-of-truth; this card is visibility only. Requires the shared Operator admin token above.
Expected Stripe event id (optional)
idle
Status
—
Safe-off
—
Billing-grade rows
—
Stamped rows
—
Pending billing-grade
—
Stripe event micros
—
Distinct event ids
—
Expected-key match
—
Errors
Warnings
Click Run reconciliation readout to fetch live state.

Next Cycle Readiness future / pending periods

Read-only. Does not call Stripe. Does not mutate DB. Not an execute surface — readiness reports what is pending; the C3.7 CLI (python -m app.operator.savings_next_cycle_readiness_cli) remains the source-of-truth for command-line operation. Requires the shared Operator admin token above.
idle
Status
—
Safe-off
—
Periods scanned
—
Ready to dry-run
—
Fully stamped
—
Mismatches
—
Pending micros
—
Stamped micros
—
Periods
Customer Period Status Pending rows Pending micros Stamped rows Stamped micros Errors / warnings
No open readiness items. Closed periods may be hidden — tick Include closed periods and re-run to see fully-stamped scopes.
Errors
Warnings
Click Run next-cycle readiness scan to fetch live state.

Recent Requests

Time Route Model Original Compressed Saved Savings Latency Status
Loading...

Cache Diagnostics what to look at first

Provider Cache Savings Anthropic prompt caching · this billing period —

Cache read tokens 0
Cache creation tokens 0
Active input tokens 0
Cache hit ratio 0.00%
Cache savings 0.00%
Provider cache reads reduce active input tokens on repeated stable prompts. Results depend on prompt reuse and provider cache eligibility.

Cost Accuracy provider-aware estimate · this billing period —

Estimated cost saved $0.00
vs no-caching baseline 0.00%
— breakdown —
Active input cost $0.00
Cache write cost $0.00
Cache read cost $0.00
Output cost $0.00
Estimated total this period $0.00
Without caching, would have been $0.00
Costs are estimated against the current published rate card. Cache writes are billed at ~125% of input rate; cache reads at ~10%. Unpriced models are excluded from totals to avoid overclaiming.

Provider Cache Savings Over Time this billing period · daily buckets

No cache activity yet in this period. Cache savings will appear here once your prompts hit the provider cache.

Prompt reuse

Repeated-prefix % (now)

Awaiting measurement data...

Simulated cache savings (full_history)

Awaiting measurement data...

Repeated-prefix trend full_history · 30d

Zur-lix Context Optimization Engine ·