Deterministic Engine
Stable, predictable results - not probabilistic guesses.
Time infrastructure for AI
AbsTime resolves natural language time expressions into deterministic time.
| " in 10 business days " Reference: 9:30 Jan 9, 2026 · New York Weekend-only counting misses national holidays like MLK Day. | " this Thanksgiving " Reference: 11:00 Sep 18, 2026 · Toronto Thanksgiving falls on a different date in Canada. | " 10 hours later " Reference: 22:00 Mar 7, 2026 · Los Angeles The clock jumps from 02:00 to 03:00 when daylight saving time begins. | ... | |
|---|---|---|---|---|
| AbsTime | Jan 26, 2026 | Oct 12, 2026 | 09:00 Mar 8, 2026 | |
| GPT-5.4 | Jan 23, 2026 | Nov 26, 2026 | 08:00 Mar 8, 2026 | |
| Claude Opus 4.6 | Jan 23, 2026 | Nov 26, 2026 | 08:00 Mar 8, 2026 | |
| Grok 4.1 | Jan 23, 2026 | Nov 26, 2026 | 08:00 Mar 8, 2026 |
Why AbsTime works where LLMs fail
LLMs guess time. AbsTime computes it.
Context-Aware
Resolves time expressions within the correct context.
Temporal Knowledge
Business days, public holidays, timezones, DST, regional calendars and more.
Guardrails
Prevents silent failures from ambiguous inputs.
Get time right without a refactor
A small change in your stack, a big improvement in user trust.
Input
" Schedule a meeting with Lily 10 hours later "
Time context
2026-03-07 22:00:00 (America/Los_Angeles)
LLM orchestration
Intent recognition
schedule_meeting()
Time extraction
10 hours later
LLM Time Resolver
2026-03-08 08:00:00
Execution
schedule_meeting(
time=2026-03-08 08:00:00,
...
)
time=2026-03-08 08:00:00,
...
)
DST starts in this interval, so only 9 hours pass.
Input
" Schedule a meeting with Lily 10 hours later "
Time context
2026-03-07 22:00:00 (America/Los_Angeles)
LLM orchestration
Intent recognition
schedule_meeting()
Time extraction
10 hours later
Execution
result = AbsTime().resolve("10 hours later", ...)
schedule_meeting(
time=2026-03-08 09:00:00,
...
)
time=2026-03-08 09:00:00,
...
)
What to know before integration
Quick answers for common integration and usage questions.
Scope
What to expect from AbsTime.
Pricing and usage
How pricing, access, and scale work.
Feedback and support
How to report issues and get help.
Does AbsTime accept full sentences?
No. AbsTime is designed to minimize unnecessary data exposure and protect user privacy. It expects clean time expressions so sensitive context stays out of the request.
How do I prepare the clean time expression?
Use a general-purpose LLM (for example, GPT, Claude, or Gemini) with an extraction prompt to produce a clean English time expression before sending it to AbsTime.
Can I use AbsTime with non-English user input?
Yes. Prompt the LLM in the time expression extraction step. The model should return the expression in English instead of the original language.
Does AbsTime resolve time ranges?
No. AbsTime is built for atomic point-in-time resolution. If you need a range, split it into two point-in-time expressions and resolve them separately with AbsTime.
How do I get the free tier?
Create an account and you will be automatically enrolled in the free tier.
What is the quota for the free tier?
The free tier includes 10,000 requests per day.
What happens if I reach the free tier limit?
Requests above the free tier limit are rejected. You can wait for the daily reset or contact us to upgrade.
Can my access be reduced or suspended?
Yes. Abusive or clearly inappropriate usage may lead to reduced access or suspension.
What should I do if a result looks wrong?
Send an email to [email protected] with the request ID or the input and output details to reproduce the result.
How soon will I hear back after I report an issue?
You will hear back from us within 24 hours after reporting an issue.
Can I get technical support during integration?
Yes. Email [email protected] if you need technical help during integration.