feat: V2 parse & extract SDK support (client.v2)

README.md

@@ -221,6 +221,105 @@ for job in response.jobs:
     print(f"Job {job.job_id}: {job.status}")
 ```
 
+## V2 API
+
+`client.v2` is a new, **additive** sub-client for LandingAI's next-generation ADE gateway. It does not replace or change anything about the V1 usage above -- `client.parse`, `client.extract`, `client.parse_jobs`, `client.extract_jobs`, etc. all keep working exactly as documented. Use `client.v2.*` when you want the newer parse/extract surface.
+
+The V2 gateway lives on its own host (`aide.[env].landing.ai`), separate from the V1 host (`api.va.[env].landing.ai`). Select the environment the same way as V1, via the `environment` argument or the `LANDINGAI_ADE_ENVIRONMENT` env var:
+
+```python
+import os
+from landingai_ade import LandingAIADE
+
+client = LandingAIADE(
+    apikey=os.environ.get("VISION_AGENT_API_KEY"),
+    # one of "production" (default), "eu", "staging", "dev"
+    # can also be set via the LANDINGAI_ADE_ENVIRONMENT env var instead of passing it here
+    environment="staging",
+)
+```
+
+### Sync parse and extract
+
+```python
+from pathlib import Path
+from landingai_ade import LandingAIADE
+
+client = LandingAIADE()
+
+parse_result = client.v2.parse(
+    document=Path("path/to/file.pdf"),
+    save_to="./output_folder",  # optional: saves as {input_file}_parse_output.json
+)
+print(parse_result.markdown)
+
+extract_result = client.v2.extract(
+    schema='{"type": "object", "properties": {"total": {"type": "string"}}}',
+    markdown=parse_result.markdown,
+    strict=False,  # prune unsupported schema fields instead of raising a 422
+)
+print(extract_result.extraction)
+```
+
+### Passing a pydantic model as the extract schema
+
+`schema` accepts a pydantic `BaseModel` subclass directly -- no need to convert it yourself:
+
+```python
+from pydantic import BaseModel, Field
+from landingai_ade import LandingAIADE
+
+
+class Invoice(BaseModel):
+    total: str = Field(description="Invoice grand total")
+
+
+client = LandingAIADE()
+result = client.v2.extract(schema=Invoice, markdown_url="https://example.com/doc.md")
+print(result.extraction)
+```
+
+### Async jobs and `wait()`
+
+For large documents, create a job and poll it, or block until it finishes with `.wait()`:
+
+```python
+from pathlib import Path
+from landingai_ade import LandingAIADE
+
+client = LandingAIADE()
+
+job = client.v2.parse_jobs.create(document=Path("path/to/large_file.pdf"), priority="priority")
+print(job.job_id, job.status)
+
+# Block until the job is terminal (raises JobWaitTimeoutError / JobFailedError on failure paths)
+done = client.v2.parse_jobs.wait(job.job_id, timeout=600, raise_on_failure=True)
+if done.result is not None:
+    print(done.result.markdown[:200])
+
+# client.v2.extract_jobs.{create,get,list,wait} mirror the same shape for extract jobs.
+```
+
+`parse_jobs.create` and `extract_jobs.create` both return a normalized `Job` -- one shape shared by parse and extract jobs, even though their upstream envelopes differ. Use `job.raw` to reach any field not surfaced on the typed model.
+
+### Uploading a file for `markdown_ref`
+
+`client.v2.files.upload` stages bytes on the ADE data plane and returns a `file_ref` you can pass as `markdown_ref` to extract:
+
+```python
+from pathlib import Path
+from landingai_ade import LandingAIADE
+
+client = LandingAIADE()
+
+file_ref = client.v2.files.upload(file=Path("path/to/file.md"))
+result = client.v2.extract(schema={"type": "object", "properties": {}}, markdown_ref=file_ref)
+```
+
+`client.v2.parse`, `client.v2.extract`, and their async counterparts also accept `save_to`, with the same auto-naming behavior as the V1 methods above.
+
+The async client mirrors this entire surface: `AsyncLandingAIADE().v2.parse(...)`, `await client.v2.parse_jobs.wait(...)`, etc.
+
 ## Async usage
 
 Simply import `AsyncLandingAIADE` instead of `LandingAIADE` and use `await` with each API call:

api.md

@@ -59,3 +59,63 @@ Methods:
 - <code title="post /v1/ade/extract/jobs">client.extract_jobs.<a href="./src/landingai_ade/resources/extract_jobs.py">create</a>(\*\*<a href="src/landingai_ade/types/extract_job_create_params.py">params</a>) -> <a href="./src/landingai_ade/types/extract_job_create_response.py">ExtractJobCreateResponse</a></code>
 - <code title="get /v1/ade/extract/jobs">client.extract_jobs.<a href="./src/landingai_ade/resources/extract_jobs.py">list</a>(\*\*<a href="src/landingai_ade/types/extract_job_list_params.py">params</a>) -> <a href="./src/landingai_ade/types/extract_job_list_response.py">ExtractJobListResponse</a></code>
 - <code title="get /v1/ade/extract/jobs/{job_id}">client.extract_jobs.<a href="./src/landingai_ade/resources/extract_jobs.py">get</a>(job_id) -> <a href="./src/landingai_ade/types/extract_job_get_response.py">ExtractJobGetResponse</a></code>
+
+# V2
+
+The `client.v2` sub-client targets LandingAI's next-generation ADE gateway, which lives on its own host (`aide.[env].landing.ai`) rather than the V1 host (`api.va.[env].landing.ai`). It is **additive**: `client.v2.*` is a separate surface from the top-level `client.*` (V1) methods documented above, and using it does not change any V1 behavior. See the [README](README.md#v2-api) for environment selection and usage examples.
+
+`client.v2.parse_jobs` and `client.v2.extract_jobs` both return a single, unified <a href="./src/landingai_ade/types/v2/job.py">`Job`</a> shape, even though the underlying parse/extract job envelopes differ upstream -- `Job.raw` retains the full original envelope as an escape hatch for any field not surfaced on the typed model.
+
+Types:
+
+```python
+from landingai_ade.types.v2 import (
+    Job,
+    JobError,
+    JobStatus,
+    V2ExtractMetadata,
+    V2ExtractResult,
+    V2FileUploadResponse,
+    V2ParseBilling,
+    V2ParseMetadata,
+    V2ParseResponse,
+)
+```
+
+- <code><a href="./src/landingai_ade/types/v2/job.py">Job</a></code> -- unified job shape: `job_id`, `status` (<code><a href="./src/landingai_ade/types/v2/job.py">JobStatus</a></code>: `pending` / `processing` / `completed` / `failed` / `cancelled`), `created_at`, `completed_at`, `progress`, `result` (a `V2ParseResponse` for parse jobs, a `V2ExtractResult` for extract jobs, or `None` until completion), `error` (<code><a href="./src/landingai_ade/types/v2/job.py">JobError</a></code>), `raw` (the full original envelope as a `dict`), and the `.is_terminal` property.
+- <code><a href="./src/landingai_ade/types/v2/parse_response.py">V2ParseResponse</a></code> -- `markdown`, `structure`, `grounding`, `metadata` (<code><a href="./src/landingai_ade/types/v2/parse_response.py">V2ParseMetadata</a></code>, which nests <code><a href="./src/landingai_ade/types/v2/parse_response.py">V2ParseBilling</a></code>). Loosely typed pending a published gateway schema; unknown fields are retained.
+- <code><a href="./src/landingai_ade/types/v2/extract_response.py">V2ExtractResult</a></code> -- `extraction`, `extraction_metadata`, `markdown`, `metadata` (<code><a href="./src/landingai_ade/types/v2/extract_response.py">V2ExtractMetadata</a></code>).
+- <code><a href="./src/landingai_ade/types/v2/file_upload_response.py">V2FileUploadResponse</a></code> -- `file_ref`.
+
+Methods:
+
+- <code title="post /v2/parse">client.v2.<a href="./src/landingai_ade/resources/v2/v2.py">parse</a>(\*, document=..., document_url=..., model=..., options=..., password=..., save_to=...) -> <a href="./src/landingai_ade/types/v2/parse_response.py">V2ParseResponse</a></code>
+
+  Synchronous parse. Provide exactly one of `document` (file) or `document_url`. Returns a `V2ParseResponse` on both full success (HTTP 200) and partial success (HTTP 206, where `result.metadata.failed_pages` lists unparsed pages). Raises `V2SyncTimeoutError` (from `landingai_ade.lib.v2_errors`) on a 504; use `parse_jobs` for long-running documents.
+
+- <code title="post /v2/parse/jobs">client.v2.parse_jobs.<a href="./src/landingai_ade/resources/v2/parse.py">create</a>(\*, document=..., document_url=..., model=..., options=..., password=..., output_save_url=..., priority=...) -> <a href="./src/landingai_ade/types/v2/job.py">Job</a></code>
+- <code title="get /v2/parse/jobs/{job_id}">client.v2.parse_jobs.<a href="./src/landingai_ade/resources/v2/parse.py">get</a>(job_id) -> <a href="./src/landingai_ade/types/v2/job.py">Job</a></code>
+- <code title="get /v2/parse/jobs">client.v2.parse_jobs.<a href="./src/landingai_ade/resources/v2/parse.py">list</a>(\*, page=..., page_size=..., status=...) -> JobList[<a href="./src/landingai_ade/types/v2/job.py">Job</a>]</code>
+- <code>client.v2.parse_jobs.<a href="./src/landingai_ade/resources/v2/parse.py">wait</a>(job_id, \*, timeout=600, poll_interval=None, raise_on_failure=False) -> <a href="./src/landingai_ade/types/v2/job.py">Job</a></code>
+
+  Blocks, polling `.get(job_id)` with exponential backoff, until the job reaches a terminal status. Raises `JobWaitTimeoutError` if `timeout` seconds elapse first, and `JobFailedError` if `raise_on_failure=True` and the job ends `failed`/`cancelled`.
+
+- <code title="post /v2/extract">client.v2.<a href="./src/landingai_ade/resources/v2/v2.py">extract</a>(\*, schema, markdown=..., markdown_ref=..., markdown_url=..., model=..., strict=..., idempotency_key=..., save_to=...) -> <a href="./src/landingai_ade/types/v2/extract_response.py">V2ExtractResult</a></code>
+
+  Synchronous extract. `schema` accepts a pydantic `BaseModel` subclass, a `dict`, or a JSON-encoded string -- all are coerced to a JSON Schema object. Provide exactly one of `markdown`, `markdown_ref` (e.g. from `client.v2.files.upload`), or `markdown_url`. `strict=True` rejects schemas with unsupported fields (HTTP 422) instead of silently pruning them. Raises `V2SyncTimeoutError` on a 504; use `extract_jobs` for long-running documents.
+
+- <code title="post /v2/extract/jobs">client.v2.extract_jobs.<a href="./src/landingai_ade/resources/v2/extract.py">create</a>(\*, schema, markdown=..., markdown_ref=..., markdown_url=..., model=..., strict=..., idempotency_key=..., priority=...) -> <a href="./src/landingai_ade/types/v2/job.py">Job</a></code>
+- <code title="get /v2/extract/jobs/{job_id}">client.v2.extract_jobs.<a href="./src/landingai_ade/resources/v2/extract.py">get</a>(job_id) -> <a href="./src/landingai_ade/types/v2/job.py">Job</a></code>
+- <code title="get /v2/extract/jobs">client.v2.extract_jobs.<a href="./src/landingai_ade/resources/v2/extract.py">list</a>(\*, page=..., page_size=..., status=...) -> JobList[<a href="./src/landingai_ade/types/v2/job.py">Job</a>]</code>
+- <code>client.v2.extract_jobs.<a href="./src/landingai_ade/resources/v2/extract.py">wait</a>(job_id, \*, timeout=600, poll_interval=None, raise_on_failure=False) -> <a href="./src/landingai_ade/types/v2/job.py">Job</a></code>
+
+  Same polling/timeout semantics as `parse_jobs.wait`. Extract jobs have no `cancelled` status, so `raise_on_failure` only ever triggers on `failed`.
+
+- <code title="post /v1/files">client.v2.files.<a href="./src/landingai_ade/resources/v2/files.py">upload</a>(\*, file) -> str</code>
+
+  Stages a file's bytes on the ADE data plane and returns a `file_ref` string, which can be passed as `markdown_ref` to `client.v2.extract`/`client.v2.extract_jobs.create`. Served on the ADE host under `/v1/files` (not `/v2/...`). Raises `LandingAiadeError` if the response has no `file_ref`.
+
+Notes:
+
+- `parse_jobs.list` / `extract_jobs.list` both return a `JobList` (a `list[Job]` subclass) carrying pagination metadata: `.has_more`, `.org_id`, `.page`, `.page_size`.
+- All `client.v2.*` methods accept the usual `extra_headers`, `extra_query`, `extra_body`, and `timeout` overrides; sync methods additionally accept `save_to` (parse/extract only, not the job-creation methods) to write the response to disk, mirroring V1's `save_to`.

examples/v2_extract.py

@@ -0,0 +1,48 @@
+#!/usr/bin/env -S rye run python
+"""Runnable example: extract structured data with the V2 (client.v2) sub-client,
+using a pydantic model directly as the extraction schema.
+
+This targets the ADE gateway host (aide.[env].landing.ai), separate from the
+V1 host used by client.extract(). It is purely additive -- V1 usage elsewhere
+in this SDK is unaffected.
+
+Set VISION_AGENT_API_KEY (and optionally LANDINGAI_ADE_ENVIRONMENT) before running:
+
+    export VISION_AGENT_API_KEY="My Apikey"
+    export LANDINGAI_ADE_ENVIRONMENT="staging"  # optional; defaults to "production"
+    ./examples/v2_extract.py
+"""
+
+from __future__ import annotations
+
+from pydantic import Field, BaseModel
+
+from landingai_ade import LandingAIADE
+from landingai_ade.types.v2 import V2ExtractResult
+
+
+class Invoice(BaseModel):
+    total: str = Field(description="Invoice grand total")
+    vendor: str = Field(description="Name of the vendor issuing the invoice")
+
+
+# apikey is read from VISION_AGENT_API_KEY; environment from LANDINGAI_ADE_ENVIRONMENT
+# (or pass apikey=..., environment=... explicitly).
+client = LandingAIADE()
+
+# `schema` accepts a pydantic BaseModel subclass directly -- it's coerced to a
+# JSON Schema dict for you. A dict or a JSON-encoded string also work.
+result = client.v2.extract(
+    schema=Invoice,
+    markdown_url="https://example.com/invoice.md",
+    strict=False,  # prune unsupported schema fields instead of raising a 422
+)
+print(result.extraction)
+
+# Long-running extractions can instead go through the async jobs route:
+job = client.v2.extract_jobs.create(schema=Invoice, markdown_url="https://example.com/invoice.md")
+done = client.v2.extract_jobs.wait(job.job_id, timeout=600)
+if isinstance(done.result, V2ExtractResult):
+    print(done.result.extraction)
+else:
+    print(f"No result available (error={done.error})")

examples/v2_parse.py

@@ -0,0 +1,43 @@
+#!/usr/bin/env -S rye run python
+"""Runnable example: parse a document with the V2 (client.v2) sub-client.
+
+This targets the ADE gateway host (aide.[env].landing.ai), separate from the
+V1 host used by client.parse(). It is purely additive -- V1 usage elsewhere in
+this SDK is unaffected.
+
+Set VISION_AGENT_API_KEY (and optionally LANDINGAI_ADE_ENVIRONMENT) before running:
+
+    export VISION_AGENT_API_KEY="My Apikey"
+    export LANDINGAI_ADE_ENVIRONMENT="staging"  # optional; defaults to "production"
+    ./examples/v2_parse.py
+"""
+
+from __future__ import annotations
+
+from pathlib import Path
+
+from landingai_ade import LandingAIADE
+from landingai_ade.types.v2 import V2ParseResponse
+
+# apikey is read from VISION_AGENT_API_KEY; environment from LANDINGAI_ADE_ENVIRONMENT
+# (or pass apikey=..., environment=... explicitly).
+client = LandingAIADE()
+
+document_path = Path(__file__).parent / "sample.pdf"
+
+# Create an async parse job for a (potentially large) document.
+job = client.v2.parse_jobs.create(document=document_path, priority="priority")
+print(f"Created parse job {job.job_id} (status={job.status})")
+
+# Block until the job reaches a terminal state.
+done = client.v2.parse_jobs.wait(job.job_id, timeout=600)
+print(f"Job finished with status={done.status}")
+
+if isinstance(done.result, V2ParseResponse):
+    print(done.result.markdown[:200] if done.result.markdown else None)
+else:
+    print(f"No result available (error={done.error})")
+
+# For smaller documents you can skip the job/wait dance entirely:
+sync_result = client.v2.parse(document=document_path)
+print(sync_result.markdown[:200] if sync_result.markdown else None)