Send your question
Submit the question you were about to put to a GPT.
Hesperia API
Better questions before you ask a GPT.
A poorly framed question gets a poor answer from any GPT. Hesperia reads your question first and returns a deterministic structural analysis of how it is built: area and family scores and a metrics bundle. You use it to reframe the question, then send it to whichever GPT you prefer.
Deterministic - No LLM call - Same question, same advice
How it works
Three Steps
Hesperia does not run, relay or answer your GPT call. It sits before it.
Get the analysis
The engine scores the structure of the question and returns the result as structured data.
Ask your GPT
Send your reframed question to any GPT you like. Hesperia is not in that path.
What you get back.
A structural analysis of how the question is built, as structured data you can act on. Hesperia never sees or answers your GPT call, so it works with any model and adds no dependency on one. It is deterministic engine compute: the same question always returns the same analysis, so results can be checked and repeated.
Modules
Default And TFP
Every call is engine compute only. The same question always gives the same analysis.
Default
The base analysis chain. It reads a question and scores its structure in four stages.
- Dictionary A: tokens to concept hits
- Dictionary B: concept hits to family and area scores
- Topology: area scores normalized
- Metrics: a metrics bundle for the question
TFP
An add-on that runs on top of Default. It reads the question on four labels and lists conflicts between them.
Usage
Commands
Copy, paste, run. Your API address and token are provided with your access.
1. Set your address and token
Use the values you received when onboarded. The token is secret: keep it out of shared code and chat.
export HESPERIA_API_URL="https://<your-api-host>" export HESPERIA_API_TOKEN="<your-token>"
2. Check access
Returns 200 with a valid token and 401 without.
curl -H "Authorization: Bearer $HESPERIA_API_TOKEN" "$HESPERIA_API_URL/health"
3. Analyze one question
Send the question you were about to put to a GPT. mode is standard (default) or reflection. Maximum 10,000 characters.
curl -X POST "$HESPERIA_API_URL/analyze" \
-H "Authorization: Bearer $HESPERIA_API_TOKEN" \
-H "Content-Type: application/json" \
-d '{"text": "Why does our team keep missing deadlines?", "mode": "standard"}'
4. Compare two questions
curl -X POST "$HESPERIA_API_URL/compare" \
-H "Authorization: Bearer $HESPERIA_API_TOKEN" \
-H "Content-Type: application/json" \
-d '{"text1": "Why do we miss deadlines?", "text2": "What blocks delivery in our team?"}'
5. Analyze many questions at once
Up to 100 questions per call.
curl -X POST "$HESPERIA_API_URL/batch-analyze" \
-H "Authorization: Bearer $HESPERIA_API_TOKEN" \
-H "Content-Type: application/json" \
-d '{"texts": ["First question", "Second question"]}'
From a program
The same call from Python. Machines and people use the same endpoints.
import os, requests
r = requests.post(
os.environ["HESPERIA_API_URL"] + "/analyze",
headers={"Authorization": "Bearer " + os.environ["HESPERIA_API_TOKEN"]},
json={"text": "Why does our team keep missing deadlines?"},
timeout=30,
)
r.raise_for_status()
print(r.json())
Limits and errors
60 requests per minute per customer. 401 missing or invalid token, 429 too many requests, 422 invalid body. Then take the result, reframe your question, and send it to the GPT of your choice.
Access
A Real API Over HTTPS
Hesperia is offered as a service API, not a web form.
Dedicated API host
Served over HTTPS from its own API subdomain, separate from this website. The address is provided with your access.
Credentials per customer
Each customer receives their own access credentials when onboarded.
Pay per question
Usage is counted per question analysed. Pricing and licence terms on request.
Access and terms
Ask For Access.
Pricing and licence terms are provided on request.