mirror of
https://github.com/BerriAI/litellm.git
synced 2026-11-17 09:34:48 +00:00
fix(ocr): read Azure Document Intelligence query options from extra_body
This commit is contained in:
parent
6cdb9959df
commit
84898ff2e5
2 changed files with 67 additions and 4 deletions
|
|
@ -72,6 +72,8 @@ async fn pages_features_and_extra_options_map_to_the_analyze_call() {
|
|||
#[case(json!({"features": "languages&pages=1"}), Error::Features)]
|
||||
#[case(json!({"req_format": "azure"}), Error::RequestFormat)]
|
||||
#[case(json!({"outputContentFormat": "html"}), Error::OutputContentFormat)]
|
||||
#[case(json!({"extra_body": {"outputContentFormat": "html"}}), Error::OutputContentFormat)]
|
||||
#[case(json!({"extra_body": {"pages": [-1]}}), Error::Pages(PagesFault::Negative))]
|
||||
#[tokio::test]
|
||||
async fn invalid_pages_features_and_format_are_rejected_before_sending(
|
||||
#[case] options: Value,
|
||||
|
|
@ -263,6 +265,51 @@ async fn the_requested_output_content_format_reaches_azure_in_its_query_spelling
|
|||
assert_eq!(response.pages[0].markdown, "hello");
|
||||
}
|
||||
|
||||
#[rstest]
|
||||
#[case::output_content_format(
|
||||
json!({"extra_body": {"outputContentFormat": "markdown"}}),
|
||||
[None, None, Some("markdown"), Some("unicodeCodePoint")]
|
||||
)]
|
||||
#[case::pages_and_features(
|
||||
json!({"extra_body": {"pages": "2", "features": ["languages"]}}),
|
||||
[Some("2"), Some("languages"), None, None]
|
||||
)]
|
||||
#[case::over_the_top_level(
|
||||
json!({"pages": [0], "outputContentFormat": "text", "extra_body": {"pages": "2-3", "outputContentFormat": "markdown"}}),
|
||||
[Some("2-3"), None, Some("markdown"), Some("unicodeCodePoint")]
|
||||
)]
|
||||
#[case::dropped(
|
||||
json!({"extra_body": {"outputContentFormat": "html"}, "drop_params": true}),
|
||||
[None, None, None, None]
|
||||
)]
|
||||
#[tokio::test]
|
||||
async fn azure_query_params_in_extra_body_reach_the_analyze_query(
|
||||
#[case] options: Value,
|
||||
#[case] expected: [Option<&str>; 4],
|
||||
) {
|
||||
let upstream = upstream([json_response(json!({
|
||||
"status": "succeeded",
|
||||
"analyzeResult": {"pages": [{"pageNumber": 1, "lines": [{"content": "hello"}]}]}
|
||||
}))])
|
||||
.await;
|
||||
|
||||
perform(ocr_request(LAYOUT_MODEL, &upstream.uri(), options))
|
||||
.await
|
||||
.unwrap();
|
||||
|
||||
let sent = only_request(&upstream).await;
|
||||
assert_eq!(
|
||||
[
|
||||
"pages",
|
||||
"features",
|
||||
"outputContentFormat",
|
||||
"stringIndexType"
|
||||
]
|
||||
.map(|name| sent.query(name)),
|
||||
expected.map(|value| value.map(String::from))
|
||||
);
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn native_format_normalizes_pages_and_keeps_the_provider_response() {
|
||||
let operation = json!({
|
||||
|
|
|
|||
|
|
@ -38,6 +38,7 @@ const AZURE_DI_API_KEY_ENV: &str = "AZURE_DOCUMENT_INTELLIGENCE_API_KEY";
|
|||
const AZURE_DI_ENDPOINT_ENV: &str = "AZURE_DOCUMENT_INTELLIGENCE_ENDPOINT";
|
||||
const OUTPUT_CONTENT_FORMAT_PARAM: &str = "outputContentFormat";
|
||||
const OUTPUT_CONTENT_FORMAT_ALIAS: &str = "output_content_format";
|
||||
const QUERY_PARAMS: [&str; 3] = ["pages", "features", OUTPUT_CONTENT_FORMAT_PARAM];
|
||||
const CODE_POINT_STRING_INDEX: &str = "unicodeCodePoint";
|
||||
const STRING_INDEX_TYPE_FIELD: &str = "stringIndexType";
|
||||
const MARKDOWN_MODEL_ID: &str = "prebuilt-layout";
|
||||
|
|
@ -217,13 +218,14 @@ impl BaseOcrConfig for AzureDocumentIntelligenceOcrConfig {
|
|||
non_default_params: &CallArguments,
|
||||
model: &str,
|
||||
) -> Result<DocumentIntelligenceParams, Error> {
|
||||
let output_content_format = requested_output_content_format(non_default_params)?;
|
||||
let params = with_query_params_from_extra_body(non_default_params);
|
||||
let output_content_format = requested_output_content_format(¶ms)?;
|
||||
let slices_markdown_pages = output_content_format == Some(OutputContentFormat::Markdown)
|
||||
&& response_format(non_default_params)? != OcrResponseFormat::Native
|
||||
&& response_format(¶ms)? != OcrResponseFormat::Native
|
||||
&& model_id(model).is_ok_and(|id| id == MARKDOWN_MODEL_ID);
|
||||
Ok(DocumentIntelligenceParams {
|
||||
pages: normalize_pages_param(non_default_params.get("pages"))?,
|
||||
features: normalize_features_param(non_default_params.get("features"))?,
|
||||
pages: normalize_pages_param(params.get("pages"))?,
|
||||
features: normalize_features_param(params.get("features"))?,
|
||||
output_content_format,
|
||||
string_index_type: slices_markdown_pages.then_some(CODE_POINT_STRING_INDEX),
|
||||
})
|
||||
|
|
@ -404,6 +406,20 @@ fn normalize_features_param(features: Option<&Value>) -> Result<Option<String>,
|
|||
Ok(Some(normalized.join(",")))
|
||||
}
|
||||
|
||||
fn with_query_params_from_extra_body(arguments: &CallArguments) -> CallArguments {
|
||||
let overrides = arguments
|
||||
.get("extra_body")
|
||||
.and_then(Value::as_object)
|
||||
.into_iter()
|
||||
.flatten()
|
||||
.filter(|(name, _)| QUERY_PARAMS.contains(&name.as_str()));
|
||||
arguments
|
||||
.iter()
|
||||
.chain(overrides)
|
||||
.map(|(name, value)| (name.clone(), value.clone()))
|
||||
.collect()
|
||||
}
|
||||
|
||||
fn requested_output_content_format(
|
||||
params: &CallArguments,
|
||||
) -> Result<Option<OutputContentFormat>, Error> {
|
||||
|
|
|
|||
Loading…
Add table
Reference in a new issue