
Cloud model service for developers...
Before going live, create a key and enable a chat service in the same project, then send requests from the sample on the service page. Knowledge injection happens inside a service marked as custom: add question-answer pairs or upload a file and then inject it. Citations are collected in a non-streaming result, and during streaming they appear where the source is used. Reports break usage into request count, latency, and spend. Load balancing and failover sit on the platform, while you still keep the key private. A private model needs its own hosting service and model details; uploading a document is not the same step. It fits teams with an internal system who want several modalities behind one project and a usage report.
SophNet is a cloud model service for developers inside an organization. It connects text, speech, and image models to an existing application. The docs say you create an API key and enable a service in a project first. Calls can use a Python package or REST, and the chat API follows a common completions format. The console reports request counts, latency, and spend, and the docs describe load balancing, failover, and elastic capacity. Knowledge can be added as a question-answer table or as uploaded files, and answers can include citation sources. A model trained elsewhere can be registered through the console model-hosting service.
Difficulty: Intermediate
Enable a service in a project
Add a service from the project overview and choose a model type. The key must belong to the same project as that service before the sample call works.
Inject Q&A and documents
In a custom service you can add question-answer pairs or upload a knowledge file and inject it. The response can include the citation sources used in that answer.
Usage and latency reports
The console totals request count, latency, and spend so you can see which enabled service is called most and whether responses slow down.
Load balancing and failover
The docs say the platform provides load balancing, failover, and elastic capacity for request changes across business periods.
Private model hosting
Choose model hosting in the console, enter the details of a model you already trained, and confirm it so the model can be called as a service.
Add chat to an internal system
Developers add text chat to an existing internal backend with a project key, instead of running a separate model stack for each app.
Answer from business documents
Inject product Q&A or manuals into a custom service so answers carry citation sources that internal reviewers can check.
Host a model you trained
After training elsewhere, the team creates a hosting service in the console and calls it from internal APIs with the project key.
The docs say to open a project, create a key, and enable a service, with the key and the model service in the same project. After the service type is set, the service detail page shows a call sample, and the credential goes in the request header. Until those steps are done, the API is not opened the way the docs describe.
No. Knowledge injection adds question-answer pairs or uploads documents inside a custom service so answers can cite sources. Model hosting registers a private model you already trained as a service, by adding a service and choosing model hosting. The docs keep the two paths separate, and uploading a manual does not by itself become a hosted model.
Real reviews and feedback from users