haystack-salesforce-npsp 0.1.0

Haystack 2.x component that ingests Salesforce Nonprofit Success Pack (NPSP) donor data as Documents for RAG.

Haystack — Salesforce NPSP

A Haystack 2.x component that ingests Salesforce Nonprofit Success Pack (NPSP) donor data as Documents for RAG pipelines — no Airbyte, no CDK.

Unlike a raw Salesforce connector, this is NPSP-aware: it understands npo02__ and npsp__ field prefixes and emits human-readable donor summaries (giving totals, gift history, engagement) instead of raw JSON.

Installation

pip install haystack-salesforce-npsp

Credentials

export SF_USERNAME="[e-mail]"
export SF_PASSWORD="your_password"
export SF_TOKEN="your_security_token"

Usage

from haystack import Pipeline
from haystack.components.embedders import SentenceTransformersDocumentEmbedder
from haystack.document_stores.in_memory import InMemoryDocumentStore
from haystack.components.writers import DocumentWriter
from haystack_salesforce_npsp import SalesforceNPSPFetcher

store = InMemoryDocumentStore()

indexing = Pipeline()
indexing.add_component("fetch", SalesforceNPSPFetcher(
    soql_filter="npo02__TotalOppAmount__c > 5000",
    limit=500,
))
indexing.add_component("embed", SentenceTransformersDocumentEmbedder())
indexing.add_component("write", DocumentWriter(document_store=store))
indexing.connect("fetch.documents", "embed.documents")
indexing.connect("embed.documents", "write.documents")

indexing.run({})

Each Document has:

• .content — narrative donor profile (name, giving summary, recent gifts)
• .meta — structured donor fields (donor_id, total_gift_amount, gift_count, affiliation, source="salesforce_npsp", …)

Affinity scores

Pass a callable (meta: dict) -> float as affinity_score_fn to inject an ML-derived donor propensity score into every document's metadata at fetch time.

SalesforceNPSPFetcher(affinity_score_fn=lambda m: min(50 + m["total_gift_amount"] / 5000, 100))

Note: because affinity_score_fn is an arbitrary callable, a fetcher configured with one is not YAML-serializable. Instantiate it in Python.