Skip to main content
What this builds. A pipeline that routes incoming queries through an LLM classifier into different response branches (greeting vs. technical question), then merges the chosen branch back into a single output. You’ll end up with. A saved AI Routing Node Example pipeline whose result output reflects whichever path the router picked for the input query. 
from vectorshift.pipeline import Pipeline
from vectorshift.request import VectorshiftApiError

PIPELINE_NAME = "AI Routing Node Example"

try:
    pipeline = Pipeline.fetch(name=PIPELINE_NAME)
    print(f"Pipeline fetched: id={pipeline.id}, branch_id={pipeline.branch_id}")
except Exception as e:
    print(f"Error fetching pipeline: {e}")
    pipeline = Pipeline.new(name=PIPELINE_NAME)
    print(f"Pipeline created: id={pipeline.id}, branch_id={pipeline.branch_id}")

try:
    # --- Input ---
    query = pipeline.add(name="query", id="query").input(input_type="string")

    # --- AI Router: classify user intent ---
    # path_0: greeting
    # path_1: technical question
    # path_else handled implicitly (no path_else on ai_routing — only numbered paths)
    router = pipeline.add(name="router", id="router").ai_routing(
        provider="openai",
        model="gpt-4o-mini",
        input_query=query.text,
        conditions=[
            "The user is greeting or saying hello",
            "The user is asking a technical question",
        ],
        outputs={"path_0": "path", "path_1": "path"},
    )

    # --- Downstream nodes gated by routing paths ---
    greeting_response = pipeline.add(
        name="greeting_response", id="greeting_response"
    ).text(
        text="Hello! How can I help you today?",
        dependencies=[router.path_0],
    )

    tech_response = pipeline.add(name="tech_response", id="tech_response").llm(
        provider="openai",
        model="gpt-4o-mini",
        stream=False,
        prompt=query.text,
        dependencies=[router.path_1],
    )

    # --- Merge the routed branches ---
    merge = pipeline.add(name="merge", id="merge").merge(
        function="first",
        type="string",
        fields=[greeting_response.text, tech_response.response],
    )

    # --- Output ---
    pipeline.add(name="result", id="result").output(
        output_type="string",
        value=merge.output,
    )

    pipeline.save()
    print(f"Pipeline saved:  id={pipeline.id}, branch_id={pipeline.branch_id}")

except VectorshiftApiError as e:
    print(f"API error: {e.status_code} {e.method} {e.endpoint}")
    print(f"Message: {e.error_message}")

​
Expected output

Pipeline created: id=..., branch_id=...
Pipeline saved:  id=..., branch_id=...
The AiRoutingNode only exposes the numbered paths you wire — there is no implicit path_else, so make sure your conditions cover the inputs you care about.

​
See also

Condition node

Deterministic branching with operators instead of an LLM.

Sub-pipelines

Compose pipelines by calling one from another.

Pipeline reference

Full method surface.