This documentation concerns the V3 of Golem.ai.
The V2 console remains accessible, a button allows you to transit from one version to another.
The contributions of the V3 are mainly simplifications:
- Remove any notion of front or target. So there is no more identification step.
- All queries are available in HTTP, Socket and Websocket without difference.
- The proposed languages on the console, in the contexts and in the archetypes, no longer contain any locale. Ex: "fr-fr" becomes "fr".
- Increase the speed of processing requests.
- Reduced startup time of a Golem.ai server (-50%).
- Reduced server RAM consumption.
- Ability to load contexts at startup (in the same way as archetypes).
- Improved protection against over-tolerance to spellcheck errors.
- Clarification of some naming.
V3 is not naturally backwards compatible with V2. A migration of contexts format and archetypes is needed.
The format of the queries is different as well. And some steps have disappeared.
The new accounts created are in V3. On the other hand, V2 accounts have a button on the interface allowing automatic migration.
Automatic migration will retrieve a transcript of all archetypes and contexts.
This will not break the servers on the V2 console. It will then manually re-create servers and associate them with contexts, and update your software that exploits the old tokens.
This manual step ensures that requests made from your programs do not suffer from migration.
Do you have a question that has not been answered here ?