Interface: RlPolicyWeightsDto
@kortexya/reasoninglayer / RlPolicyWeightsDto
Interface: RlPolicyWeightsDto
Section titled “Interface: RlPolicyWeightsDto”Defined in: src/types/rl-training.ts:108
A tenant’s persisted RL policy weights, transported as base64.
Remarks
Section titled “Remarks”Downloaded via GET /api/v1/rl/policy/weights and uploaded via
PUT /api/v1/rl/policy/weights. The payload is the name-keyed binary produced
by the neural adapter — the same bytes RlTrainRequest.persistWeights
writes to the tenant checkpoint — so a downloaded policy can be re-uploaded
(or moved across tenants/deployments) and warm-started from.
Serialized to the snake_case wire shape
({"weights_base64": "...", "size_bytes": 1024}). weightsBase64 uses the
standard base64 alphabet and sizeBytes is the decoded payload size.
Example
Section titled “Example”const weights = await client.rlTraining.getPolicyWeights();console.log(weights.sizeBytes, weights.weightsBase64.slice(0, 16));Properties
Section titled “Properties”sizeBytes
Section titled “sizeBytes”sizeBytes:
number
Defined in: src/types/rl-training.ts:112
Decoded payload size in bytes.
weightsBase64
Section titled “weightsBase64”weightsBase64:
string
Defined in: src/types/rl-training.ts:110
The serialized policy weights, base64 (standard alphabet).