Sync wasm experiment branch to 0.20 (#213)

* Update to Assistants example (#146) * Update to Assistants example * Update examples/assistants/src/main.rs update api config for consistency and secutity Co-authored-by: Himanshu Neema <[email protected]> * added assistant creation * exit, deconstruct assistant, improved readme --------- Co-authored-by: Himanshu Neema <[email protected]> * Add examples tool-call and tool-call-stream (#153) * add names (#150) * Link to openai-func-enums (#152) * Link to openai-func-enums * Link to openai-func-enums * Update async-openai/README.md --------- Co-authored-by: Himanshu Neema <[email protected]> * In memory files (#154) * Added ability to use in-memory files (Bytes, vec[u8]) * Removed unnecessary trait impls * Polished example * Spec, readme, and crate description updates (#156) * get latest spec * update description * add WASM * WASM support on experiments branch * chore: Release * Make tool choice lower case (#158) * Fix: post_form to be Sendable (#157) * changed to allow Send. * add simple tests for sendable * fix test name * chore: Release * Add support for rustls-webpki-roots (#168) * Refactor `types` module (#170) * Document `impl_from!` macro * Fix up `impl_from!` docs * Documents `impl_default!` macro * Document `impl_input!` macro * Factor out types from `assistants` module in `types` * Factor out `model` * Factor out `audio` * Factor out `image` * Factor out `file` * Factor out `fine_tune` * Factor out `moderation` * Factor out `edit` * Factor out `fine_tuning` * Factor out missed `DeleteModelResponse` into `model` * Factor out `embedding` * Factor out `chat` * Factor out `completion` and eliminate `types` * Satisfy clippy --------- Co-authored-by: Sharif Haason <[email protected]> * Sync updates from Spec (#171) * updates to doc comments and types * deprecated * update ChatCompletionFunctions to FunctionObject * More type updates * add logprobs field * update from spec * updated spec * fixes suggested by cargo clippy * add query param to list files (#172) * chore: Release * Optional model in ModifyAssistantRequest (#174) All fields (including model) are optional in OpenAI API. * update contribution guidelines (#182) * update contribution guidelines * fix link * update * consistency * Code of conduct * chore: Release * fix file test by providing query param * Added dimensions param to embedding request (#185) * chore: Release * fix: CreateTranscriptionRequest language field not convert (#188) * chore: Release * Add usage information to the run object (#195) * Updates from Spec (#196) * updates from spec * remove Edits * remove Fine-Tunes (was deprecated) * update spec * cargo fix * cargo fmt * chore: Release * Add Client::build for full customizability during instantiation (#197) * Change std::sleep to tokio's sleep (#200) * chore: Release * add support for base64 embeddings (#190) * add support for base64 embeddings * Base64Embedding is an implementation detail * feat: separate Embeddings::create_base64 method * chore: use newtype for hosting base64 decoding instead * chore: remove unused error variant * Add vision-chat example (#203) Example matches quickstart from https://platform.openai.com/docs/guides/vision It showcases struct derived from ChatCompletionRequestMessageContent * Update Audio APIs from updated spec (#202) * Implement CreateTranscriptRequest::response_granularities This PR adds support for `AudioResponseFormat::VerboseJson` and `TimestampGranularity`, including updated example code. These were defined as types before, but not fully implemented. Implements #201. * Modify transcription API to be more like spec - Rename `CreateTranscriptionRespose` to `CreateTranscriptionResponseJson` (to match API spec) - Add `CreateTranscriptionResponseVerboseJson` and `transcribe_verbose_json` - Add `transcribe_raw` for SRT output - Add `post_form_raw` - Update example code * Upgrade dependencies: Rust crates in Cargo.toml (#204) * upgrade reqwest * update reqwest-eventsource * cargo test working (#207) * fix: cargo fmt and compiler warnings fixes (#208) * cargo fmt * fix imports * chore: Release * fixed problems due to code sync * update worker dependency to resolve build issue * update test to fix test compilation issue * add conditional imports * change default of InputSource and bring back builders of file-related structs * update doc --------- Co-authored-by: Gravel Hill <[email protected]> Co-authored-by: Himanshu Neema <[email protected]> Co-authored-by: Frank Fralick <[email protected]> Co-authored-by: Sam F <[email protected]> Co-authored-by: David Weis <[email protected]> Co-authored-by: yykt <[email protected]> Co-authored-by: XTY <[email protected]> Co-authored-by: sharif <[email protected]> Co-authored-by: Sharif Haason <[email protected]> Co-authored-by: Sebastian Sosa <[email protected]> Co-authored-by: vmg-dev <[email protected]> Co-authored-by: TAO <[email protected]> Co-authored-by: turingbuilder <[email protected]> Co-authored-by: Gabriel Bianconi <[email protected]> Co-authored-by: Santhanagopalan Krishnamoorthy <[email protected]> Co-authored-by: Adrien Wald <[email protected]> Co-authored-by: Gabriel <[email protected]> Co-authored-by: Eric Kidd <[email protected]> Co-authored-by: Samuel Batissou Tiburcio <[email protected]>
64bit · Apr 10, 2024 · 580ff11 · 580ff11
1 parent 50d661f
commit 580ff11
Show file tree

Hide file tree

Showing 41 changed files with 1,101 additions and 1,776 deletions.
diff --git a/async-openai/Cargo.toml b/async-openai/Cargo.toml
@@ -1,6 +1,6 @@
 [package]
 name = "async-openai"
-version = "0.18.2"
+version = "0.20.0"
 authors = [
     "Himanshu Neema"
 ]
@@ -33,16 +33,16 @@ native-tls-vendored = ["reqwest/native-tls-vendored"]
 
 [dependencies]
 backoff = {version = "0.4.0", features = ["futures"], optional = true }
-base64 = "0.21.0"
+base64 = "0.22.0"
 futures = "0.3.26"
 rand = "0.8.5"
-reqwest = { version = "0.11.14", features = ["json", "stream", "multipart"],default-features = false }
-reqwest-eventsource = "0.5.0"
+reqwest = { version = "0.12.0", features = ["json", "stream", "multipart"],default-features = false }
+reqwest-eventsource = "0.6.0"
 serde = { version = "1.0.152", features = ["derive", "rc"] }
 serde_json = "1.0.93"
 thiserror = "1.0.38"
 tracing = "0.1.37"
-derive_builder = "0.12.0"
+derive_builder = "0.20.0"
 async-convert = "1.0.0"
 secrecy = { version = "0.8.0", features=["serde"] }
 tokio = { version = "1.25.0", features = ["fs", "macros"], optional = true }

diff --git a/async-openai/README.md b/async-openai/README.md
@@ -24,14 +24,12 @@
 - It's based on [OpenAI OpenAPI spec](https://github.com/openai/openai-openapi)
 - Current features:
   - [x] Assistants (Beta)
-  - [x] Audio (Whisper/TTS)
+  - [x] Audio
   - [x] Chat
   - [x] Completions (Legacy)
-  - [x] Edits (Deprecated)
   - [x] Embeddings
   - [x] Files
   - [x] Fine-Tuning
-  - [x] Fine-Tunes (Deprecated)
   - [x] Images
   - [x] Microsoft Azure OpenAI Service
   - [x] Models
@@ -125,7 +123,7 @@ All forms of contributions, such as new features requests, bug fixes, issues, do
 A good starting point would be to look at existing [open issues](https://github.com/64bit/async-openai/issues).
 
 To maintain quality of the project, a minimum of the following is a must for code contribution:
-- **Documented**: Primary source of doc comments is description field from OpenAPI spec.
+- **Names & Documentation**: All struct names, field names and doc comments are from OpenAPI spec. Nested objects in spec without names leaves room for making appropriate name.
 - **Tested**: Examples are primary means of testing and should continue to work. For new features supporting example is required.
 - **Scope**: Keep scope limited to APIs available in official documents such as [API Reference](https://platform.openai.com/docs/api-reference) or [OpenAPI spec](https://github.com/openai/openai-openapi/). Other LLMs or AI Providers offer OpenAI-compatible APIs, yet they may not always have full parity. In such cases, the OpenAI spec takes precedence.
 - **Consistency**: Keep code style consistent across all the "APIs" that library exposes; it creates a great developer experience.
@@ -135,9 +133,6 @@ This project adheres to [Rust Code of Conduct](https://www.rust-lang.org/policie
 ## Complimentary Crates
 - [openai-func-enums](https://github.com/frankfralick/openai-func-enums) provides procedural macros that make it easier to use this library with OpenAI API's tool calling feature. It also provides derive macros you can add to existing [clap](https://github.com/clap-rs/clap) application subcommands for natural language use of command line tools. It also supports openai's [parallel tool calls](https://platform.openai.com/docs/guides/function-calling/parallel-function-calling) and allows you to choose between running multiple tool calls concurrently or own their own OS threads.
 
-## Complimentary Crates
-- [openai-func-enums](https://github.com/frankfralick/openai-func-enums) provides procedural macros that make it easier to use this library with OpenAI API's tool calling feature. It also provides derive macros you can add to existing [clap](https://github.com/clap-rs/clap) application subcommands for natural language use of command line tools. It also supports openai's [parallel tool calls](https://platform.openai.com/docs/guides/function-calling/parallel-function-calling) and allows you to choose between running multiple tool calls concurrently or own their own OS threads.
-
 
 ## License
 

diff --git a/async-openai/src/audio.rs b/async-openai/src/audio.rs
@@ -1,9 +1,12 @@
+use bytes::Bytes;
+
 use crate::{
     config::Config,
     error::OpenAIError,
     types::{
         CreateSpeechRequest, CreateSpeechResponse, CreateTranscriptionRequest,
-        CreateTranscriptionResponse, CreateTranslationRequest, CreateTranslationResponse,
+        CreateTranscriptionResponseJson, CreateTranscriptionResponseVerboseJson,
+        CreateTranslationRequest, CreateTranslationResponse,
     },
     Client,
 };
@@ -23,12 +26,32 @@ impl<'c, C: Config> Audio<'c, C> {
     pub async fn transcribe(
         &self,
         request: CreateTranscriptionRequest,
-    ) -> Result<CreateTranscriptionResponse, OpenAIError> {
+    ) -> Result<CreateTranscriptionResponseJson, OpenAIError> {
+        self.client
+            .post_form("/audio/transcriptions", request)
+            .await
+    }
+
+    /// Transcribes audio into the input language.
+    pub async fn transcribe_verbose_json(
+        &self,
+        request: CreateTranscriptionRequest,
+    ) -> Result<CreateTranscriptionResponseVerboseJson, OpenAIError> {
         self.client
             .post_form("/audio/transcriptions", request)
             .await
     }
 
+    /// Transcribes audio into the input language.
+    pub async fn transcribe_raw(
+        &self,
+        request: CreateTranscriptionRequest,
+    ) -> Result<Bytes, OpenAIError> {
+        self.client
+            .post_form_raw("/audio/transcriptions", request)
+            .await
+    }
+
     /// Translates audio into into English.
     pub async fn translate(
         &self,

diff --git a/async-openai/src/client.rs b/async-openai/src/client.rs
@@ -14,10 +14,10 @@ use crate::{
     config::{Config, OpenAIConfig},
     error::{map_deserialization_error, OpenAIError, WrappedError},
     moderation::Moderations,
-    edit::Edits,
     file::Files,
     image::Images,
-    Chat, Completions, Embeddings, Models, FineTunes, FineTuning, Assistants, Threads, Audio};
+    Assistants, Audio, Chat, Completions, Embeddings, FineTuning, Models, Threads,
+};
 
 #[derive(Debug, Clone)]
 /// Client is a container for config, backoff and http_client
@@ -42,6 +42,21 @@ impl Client<OpenAIConfig> {
 }
 
 impl<C: Config> Client<C> {
+    /// Create client with a custom HTTP client, OpenAI config, and backoff.
+    pub fn build(
+        http_client: reqwest::Client,
+        config: C,
+        #[cfg(feature = "backoff")]
+        backoff: backoff::ExponentialBackoff,
+    ) -> Self {
+        Self {
+            http_client,
+            config,
+            #[cfg(feature = "backoff")]
+            backoff,
+        }
+    }
+
     /// Create client with [OpenAIConfig] or [crate::config::AzureConfig]
     pub fn with_config(config: C) -> Self {
         Self {
@@ -84,12 +99,6 @@ impl<C: Config> Client<C> {
         Chat::new(self)
     }
 
-    /// To call [Edits] group related APIs using this client.
-    #[deprecated(since = "0.15.0", note = "By OpenAI")]
-    pub fn edits(&self) -> Edits<C> {
-        Edits::new(self)
-    }
-
     /// To call [Images] group related APIs using this client.
     pub fn images(&self) -> Images<C> {
         Images::new(self)
@@ -105,12 +114,6 @@ impl<C: Config> Client<C> {
         Files::new(self)
     }
 
-    /// To call [FineTunes] group related APIs using this client.
-    #[deprecated(since = "0.15.0", note = "By OpenAI")]
-    pub fn fine_tunes(&self) -> FineTunes<C> {
-        FineTunes::new(self)
-    }
-
     /// To call [FineTuning] group related APIs using this client.
     pub fn fine_tuning(&self) -> FineTuning<C> {
         FineTuning::new(self)
@@ -230,6 +233,25 @@ impl<C: Config> Client<C> {
         self.execute(request_maker).await
     }
 
+    /// POST a form at {path} and return the response body
+    pub(crate) async fn post_form_raw<F>(&self, path: &str, form: F) -> Result<Bytes, OpenAIError>
+    where
+        reqwest::multipart::Form: async_convert::TryFrom<F, Error = OpenAIError>,
+        F: Clone,
+    {
+        let request_maker = || async {
+            Ok(self
+                .http_client
+                .post(self.config.url(path))
+                .query(&self.config.query())
+                .headers(self.config.headers())
+                .multipart(async_convert::TryFrom::try_from(form.clone()).await?)
+                .build()?)
+        };
+
+        self.execute_raw(request_maker).await
+    }
+
     /// POST a form at {path} and deserialize the response body
     pub(crate) async fn post_form<O, F>(&self, path: &str, form: F) -> Result<O, OpenAIError>
     where

diff --git a/async-openai/src/edit.rs b/async-openai/src/edit.rs
diff --git a/async-openai/src/embedding.rs b/async-openai/src/embedding.rs
@@ -1,7 +1,10 @@
 use crate::{
     config::Config,
     error::OpenAIError,
-    types::{CreateEmbeddingRequest, CreateEmbeddingResponse},
+    types::{
+        CreateBase64EmbeddingResponse, CreateEmbeddingRequest, CreateEmbeddingResponse,
+        EncodingFormat,
+    },
     Client,
 };
 
@@ -23,14 +26,36 @@ impl<'c, C: Config> Embeddings<'c, C> {
         &self,
         request: CreateEmbeddingRequest,
     ) -> Result<CreateEmbeddingResponse, OpenAIError> {
+        if matches!(request.encoding_format, Some(EncodingFormat::Base64)) {
+            return Err(OpenAIError::InvalidArgument(
+                "When encoding_format is base64, use Embeddings::create_base64".into(),
+            ));
+        }
+        self.client.post("/embeddings", request).await
+    }
+
+    /// Creates an embedding vector representing the input text.
+    ///
+    /// The response will contain the embedding in base64 format.
+    pub async fn create_base64(
+        &self,
+        request: CreateEmbeddingRequest,
+    ) -> Result<CreateBase64EmbeddingResponse, OpenAIError> {
+        if !matches!(request.encoding_format, Some(EncodingFormat::Base64)) {
+            return Err(OpenAIError::InvalidArgument(
+                "When encoding_format is not base64, use Embeddings::create".into(),
+            ));
+        }
+
         self.client.post("/embeddings", request).await
     }
 }
 
 #[cfg(test)]
 mod tests {
+    use crate::error::OpenAIError;
+    use crate::types::{CreateEmbeddingResponse, Embedding, EncodingFormat};
     use crate::{types::CreateEmbeddingRequestArgs, Client};
-    use crate::types::{CreateEmbeddingResponse, Embedding};
 
     #[tokio::test]
     async fn test_embedding_string() {
@@ -122,9 +147,61 @@ mod tests {
 
         assert!(response.is_ok());
 
-        let CreateEmbeddingResponse { mut data, ..} = response.unwrap();
+        let CreateEmbeddingResponse { mut data, .. } = response.unwrap();
         assert_eq!(data.len(), 1);
         let Embedding { embedding, .. } = data.pop().unwrap();
         assert_eq!(embedding.len(), dimensions as usize);
     }
+
+    #[tokio::test]
+    async fn test_cannot_use_base64_encoding_with_normal_create_request() {
+        let client = Client::new();
+
+        const MODEL: &str = "text-embedding-ada-002";
+        const INPUT: &str = "You shall not pass.";
+
+        let b64_request = CreateEmbeddingRequestArgs::default()
+            .model(MODEL)
+            .input(INPUT)
+            .encoding_format(EncodingFormat::Base64)
+            .build()
+            .unwrap();
+        let b64_response = client.embeddings().create(b64_request).await;
+        assert!(matches!(b64_response, Err(OpenAIError::InvalidArgument(_))));
+    }
+
+    #[tokio::test]
+    async fn test_embedding_create_base64() {
+        let client = Client::new();
+
+        const MODEL: &str = "text-embedding-ada-002";
+        const INPUT: &str = "CoLoop will eat the other qual research tools...";
+
+        let b64_request = CreateEmbeddingRequestArgs::default()
+            .model(MODEL)
+            .input(INPUT)
+            .encoding_format(EncodingFormat::Base64)
+            .build()
+            .unwrap();
+        let b64_response = client
+            .embeddings()
+            .create_base64(b64_request)
+            .await
+            .unwrap();
+        let b64_embedding = b64_response.data.into_iter().next().unwrap().embedding;
+        let b64_embedding: Vec<f32> = b64_embedding.into();
+
+        let request = CreateEmbeddingRequestArgs::default()
+            .model(MODEL)
+            .input(INPUT)
+            .build()
+            .unwrap();
+        let response = client.embeddings().create(request).await.unwrap();
+        let embedding = response.data.into_iter().next().unwrap().embedding;
+
+        assert_eq!(b64_embedding.len(), embedding.len());
+        for (b64, normal) in b64_embedding.iter().zip(embedding.iter()) {
+            assert!((b64 - normal).abs() < 1e-6);
+        }
+    }
 }
diff --git a/async-openai/src/file.rs b/async-openai/src/file.rs
@@ -55,6 +55,7 @@ impl<'c, C: Config> Files<'c, C> {
 }
 
 #[cfg(test)]
+#[cfg(not(feature = "wasm"))]
 mod tests {
     use crate::{types::CreateFileRequestArgs, Client};