Infratex Python SDK

Official Python client for the Infratex document intelligence API. Parse PDFs or ordered image batches, build search indexes, and generate AI-powered answers grounded in your documents.

Installation

pip install infratex

Quick start

from infratex import Infratex

client = Infratex(api_key="infratex_sk_...")

# Upload and parse a PDF
doc = client.documents.upload("report.pdf")
print(doc.id, doc.status, doc.page_count)

# Upload an ordered image batch as document pages
deck = client.documents.upload_images(["page-1.png", "page-2.png"], method="max")
print(deck.id, deck.status, deck.page_count)

# Index for search
# The SDK waits for the queued index by default.
index = client.documents.index(doc.id, method="vector")

# Search
# Searches and responses require a ready index that matches the selected method.
results = client.searches.create(
    query="revenue growth",
    method="vector",
    document_ids=[doc.id],
)
for r in results:
    print(r.score, r.content[:100])

# AI response (streamed)
for event in client.responses.create(message="Summarize the key findings", document_ids=[doc.id]):
    if event.type == "text":
        print(event.content, end="")
    elif event.type == "sources":
        print("Sources:", event.content)

Authentication

Pass your API key directly or set the INFRATEX_API_KEY environment variable:

# Explicit
client = Infratex(api_key="infratex_sk_...")

# From environment
import os
os.environ["INFRATEX_API_KEY"] = "infratex_sk_..."
client = Infratex()

Resources

Documents

# Upload
# The SDK keeps this ergonomic one-call flow even though the raw HTTP API
# now creates the document first and polls until parsing is complete.
doc = client.documents.upload("report.pdf")
doc = client.documents.upload("report.pdf", method="standard", collection_id="col-id")
doc = client.documents.upload("deck.pdf", method="max")

# Upload ordered images instead of a PDF
images = client.documents.upload_images(["page-1.png", "page-2.png"])
images = client.documents.upload_images(["page-1.png", "page-2.png"], method="max", collection_id="col-id")

# Queue-first upload if you want to manage the parse lifecycle yourself
queued = client.documents.upload("report.pdf", wait=False)
doc = client.documents.get(queued.id, wait=True)

# Queue-first image upload follows the same pattern
queued_images = client.documents.upload_images(["page-1.png", "page-2.png"], wait=False)
images = client.documents.get(queued_images.id, wait=True)

# List
docs = client.documents.list(limit=50, offset=0, collection_id="col-id")
print(docs.total)
for d in docs:
    print(d.filename)

# Get
doc = client.documents.get("doc-id")

# Download markdown
md = client.documents.markdown("doc-id")

# Delete
client.documents.delete("doc-id")

# Index
# By default this waits until the queued method-specific index reaches "indexed".
index = client.documents.index("doc-id", method="hybrid")

# Queue-first behavior if you want to manage polling yourself
queued = client.documents.index("doc-id", method="hybrid", wait=False)
indexes = client.documents.list_indexes("doc-id")
index = client.documents.get_index("doc-id", "hybrid", wait=True)

Searches

results = client.searches.create(
    query="What is the EBITDA?",
    method="vector",
    limit=5,
    document_ids=["doc-id"],
)
for r in results:
    print(r.score, r.content[:200])

Responses (streaming)

for event in client.responses.create(
    message="Summarize the report",
    method="hybrid",
    limit=5,
    document_ids=["doc-id"],
):
    if event.type == "text":
        print(event.content, end="")
    elif event.type == "sources":
        print("Sources:", event.content)
    elif event.type == "done":
        print("\n--- Done ---")

# Managed multi-turn thread with persisted scope
conv = client.conversations.create(
    title="Quarterly Analysis",
    collection_id="col-id",
)

for event in client.responses.create(
    message="How does that compare with the previous quarter?",
    method="hybrid",
    model="pro",
    conversation_id=conv.id,
):
    if event.type == "text":
        print(event.content, end="")

documents.upload(...), documents.upload_images(...), and documents.index(...) now follow the same contract: they wait by default, support queue-first control with wait=False, and expose a corresponding getter with wait=True when you want to resume later.

Use method="max" when you want the Gemini parser to preserve the same extracted text while also appending brief [visual-note: ...] lines for meaningful charts, figures, screenshots, and photos.

Collections

col = client.collections.create(name="Q3 Reports")
cols = client.collections.list()
col = client.collections.get("col-id")
client.collections.update("col-id", name="Q4 Reports")
client.collections.delete("col-id")

Conversations

conv = client.conversations.create(title="Analysis", collection_id="col-id")
convs = client.conversations.list()
conv = client.conversations.get("conv-id")  # includes messages
client.conversations.delete("conv-id")

Account & Billing

account = client.account.get()
print(account.tenant["email"])

billing = client.billing.get()
print(billing.balance_micros)

Error handling

from infratex import Infratex, InfratexError

client = Infratex(api_key="infratex_sk_...")

try:
    doc = client.documents.get("nonexistent-id")
except InfratexError as e:
    print(e.status_code)  # 404
    print(e.code)         # error code from the API
    print(str(e))         # human-readable message

Configuration

client = Infratex(
    api_key="infratex_sk_...",
    base_url="https://api.infratex.io",  # custom base URL
    timeout=60.0,                         # request timeout in seconds
)

# Use as a context manager
with Infratex(api_key="infratex_sk_...") as client:
    doc = client.documents.upload("report.pdf")

License

MIT