Add multi-label docs

[xmariachi]

Purpose of Changes and their Description

Update docs and changelog wrt multi-label

Are these changes tested and documented?

• [] If tested, please describe how. If not, why tests are not needed. -- no need: just docs change
• If documented, please describe where. If not, describe why docs are not needed. -- yes, this is docs change
• Added to Unreleased section of CHANGELOG.md? -- added here the multi-label with release

---

Summary by cubic

Docs update to explain multi-label (classification) support and how arity is decided by the topic’s on-chain output arity. Adds labeled worker/reputer config examples, request/response formats, the allora_worker_labeled_inference_request_count metric, and updates CHANGELOG with v0.14.0 and an Unreleased section.

• Migration
  • MULTI topics require LabeledInferenceEndpoint, LabeledGroundTruthEndpoint, and LabeledLossFunctionService.
  • SINGLE topics require InferenceEndpoint, GroundTruthEndpoint, and LossFunctionService.
  • The node fails fast if required endpoints for the topic’s arity are missing; endpoints for the other arity are ignored with a warning.

^{Written for commit 9ffce6d. Summary will update on new commits.}

[cubic-dev-ai]

1 issue found across 3 files

Prompt for AI agents (unresolved issues)

Check if these issues are valid — if so, understand the root cause of each and fix them. If appropriate, use sub-agents to investigate and fix each issue separately.


<file name="CHANGELOG.md">

<violation number="1" location="CHANGELOG.md:56">
P1: The diff adds a second `## Unreleased` block and a second `## v0.14.0` block to the changelog, but both sections already exist in the file (the existing `## Unreleased` sits between the template and v0.14.0, and v0.14.0 already has the multi-label entries). This produces two unreleased sections and two v0.14.0 sections. Remove the duplicate insert and, if needed, add any new unreleased entries directly into the existing `## Unreleased` section.</violation>
</file>

Architecture diagram

sequenceDiagram
    participant Node as Off-Chain Node
    participant Chain as On-Chain (allora-chain)
    participant WorkerSource as Inference Source (external)
    participant ReputerSource as Ground Truth Source (external)
    participant LossSvc as Loss Function Service

    Note over Node: Startup Phase – per topic

    Node->>Chain: Query TopicOutputArity for topicId
    Chain-->>Node: arity = SINGLE | MULTI

    alt arity == SINGLE
        Node->>Node: Check InferenceEndpoint / GroundTruthEndpoint / LossFunctionService exist
        opt LabeledInferenceEndpoint etc. also configured
            Node->>Node: Log warning, ignore multi-label endpoints
        end
        alt Required scalar endpoint missing
            Node->>Node: Fail fast (panic / error)
        end
    else arity == MULTI
        Node->>Node: Check LabeledInferenceEndpoint / LabeledGroundTruthEndpoint / LabeledLossFunctionService exist
        opt InferenceEndpoint etc. also configured
            Node->>Node: Log warning, ignore scalar endpoints
        end
        alt Required labeled endpoint missing
            Node->>Node: Fail fast (panic / error)
        end
    end

    Note over Node: Per-Epoch Worker Flow

    Note over Node: Worker – Inference
    alt arity == SINGLE
        Node->>WorkerSource: GET inference (InferenceEndpoint)
        WorkerSource-->>Node: scalar value
        Node->>Node: Increment allora_worker_inference_request_count
    else arity == MULTI
        Node->>WorkerSource: GET labeled inference (LabeledInferenceEndpoint)
        WorkerSource-->>Node: JSON array [{label, value}, …]
        Node->>Node: Increment allora_worker_labeled_inference_request_count (NEW metric)
    end
    Node->>Chain: Submit inference (scalar or labeled)

    Note over Node: Per-Epoch Reputer Flow

    Note over Node: Reputer – Ground Truth & Loss
    alt arity == SINGLE
        Node->>ReputerSource: GET ground truth (GroundTruthEndpoint)
        ReputerSource-->>Node: scalar value
        Node->>LossSvc: POST /calculate (LossFunctionService)
        LossSvc-->>Node: loss value
    else arity == MULTI
        Node->>ReputerSource: GET labeled ground truth (LabeledGroundTruthEndpoint)
        ReputerSource-->>Node: JSON array [{label, value}, …]
        Node->>LossSvc: POST /calculate (LabeledLossFunctionService)
        LossSvc-->>Node: loss value
    end
    Node->>Chain: Submit reputer value (scalar or labeled) and loss

Loading

<sub>Reply with feedback, questions, or to request a fix.

Fix all with cubic | Re-trigger cubic</sub>