From 6056573b31edb02ad152fef00046e0d8607a96cb Mon Sep 17 00:00:00 2001 From: writinwaters <93570324+writinwaters@users.noreply.github.com> Date: Fri, 18 Oct 2024 20:56:33 +0800 Subject: [PATCH] DRAFT: Updated chunk APIs (#2901) ### What problem does this PR solve? ### Type of change - [x] Documentation Update --------- Signed-off-by: Jin Hai Co-authored-by: Jin Hai --- api/http_api.md | 6 +- api/python_api_reference.md | 367 ++++++++++++++++++++++-------------- 2 files changed, 231 insertions(+), 142 deletions(-) diff --git a/api/http_api.md b/api/http_api.md index 99c8363e6f5..d5a789afddb 100644 --- a/api/http_api.md +++ b/api/http_api.md @@ -37,7 +37,7 @@ Creates a dataset. # "name": name is required and can't be duplicated. # "tenant_id": tenant_id must not be provided. # "embedding_model": embedding_model must not be provided. -# "navie" means general. +# "naive" means general. curl --request POST \ --url http://{address}/api/v1/dataset \ --header 'Content-Type: application/json' \ @@ -236,7 +236,7 @@ Updates a dataset by its id. # "chunk_count": If you update chunk_count, it can't be changed. # "document_count": If you update document_count, it can't be changed. # "parse_method": If you update parse_method, chunk_count must be 0. -# "navie" means general. +# "naive" means general. curl --request PUT \ --url http://{address}/api/v1/dataset/{dataset_id} \ --header 'Content-Type: application/json' \ @@ -247,7 +247,7 @@ curl --request PUT \ "embedding_model": "BAAI/bge-zh-v1.5", "chunk_count": 0, "document_count": 0, - "parse_method": "navie" + "parse_method": "naive" }' ``` diff --git a/api/python_api_reference.md b/api/python_api_reference.md index 9e1b49704bc..4efc29e12cd 100644 --- a/api/python_api_reference.md +++ b/api/python_api_reference.md @@ -3,10 +3,10 @@ **THE API REFERENCES BELOW ARE STILL UNDER DEVELOPMENT.** :::tip NOTE -Knowledge Base Management +Dataset Management ::: -## Create knowledge base +## Create dataset ```python RAGFlow.create_dataset( @@ -17,12 +17,12 @@ RAGFlow.create_dataset( permission: str = "me", document_count: int = 0, chunk_count: int = 0, - parse_method: str = "naive", + chunk_method: str = "naive", parser_config: DataSet.ParserConfig = None ) -> DataSet ``` -Creates a knowledge base (dataset). +Creates a dataset. ### Parameters @@ -42,38 +42,24 @@ The unique name of the dataset to create. It must adhere to the following requir Base64 encoding of the avatar. Defaults to `""` -#### description - -#### tenant_id: `str` - -The id of the tenant associated with the created dataset is used to identify different users. Defaults to `None`. - -- If creating a dataset, tenant_id must not be provided. -- If updating a dataset, tenant_id can't be changed. - #### description: `str` -The description of the created dataset. Defaults to `""`. +A brief description of the dataset to create. Defaults to `""`. #### language: `str` -The language setting of the created dataset. Defaults to `"English"`. ???????????? - -#### permission - -Specify who can operate on the dataset. Defaults to `"me"`. +The language setting of the dataset to create. Available options: -#### document_count: `int` +- `"English"` (Default) +- `"Chinese"` -The number of documents associated with the dataset. Defaults to `0`. - -#### chunk_count: `int` +#### permission -The number of data chunks generated or processed by the created dataset. Defaults to `0`. +Specifies who can operate on the dataset. You can set it only to `"me"` for now. -#### parse_method, `str` +#### chunk_method, `str` -The method used by the dataset to parse and process data. Defaults to `"naive"`. +The default parsing method of the knwoledge . Defaults to `"naive"`. #### parser_config @@ -100,19 +86,19 @@ ds = rag_object.create_dataset(name="kb_1") --- -## Delete knowledge bases +## Delete datasets ```python RAGFlow.delete_datasets(ids: list[str] = None) ``` -Deletes knowledge bases by name or ID. +Deletes datasets by name or ID. ### Parameters #### ids -The IDs of the knowledge bases to delete. +The IDs of the datasets to delete. ### Returns @@ -127,7 +113,7 @@ rag.delete_datasets(ids=["id_1","id_2"]) --- -## List knowledge bases +## List datasets ```python RAGFlow.list_datasets( @@ -140,7 +126,7 @@ RAGFlow.list_datasets( ) -> list[DataSet] ``` -Retrieves a list of knowledge bases. +Retrieves a list of datasets. ### Parameters @@ -158,7 +144,7 @@ The field by which the records should be sorted. This specifies the attribute or #### desc: `bool` -Whether the sorting should be in descending order. Defaults to `True`. +Indicates whether the retrieved datasets should be sorted in descending order. Defaults to `True`. #### id: `str` @@ -170,19 +156,19 @@ The name of the dataset to be got. Defaults to `None`. ### Returns -- Success: A list of `DataSet` objects representing the retrieved knowledge bases. +- Success: A list of `DataSet` objects representing the retrieved datasets. - Failure: `Exception`. ### Examples -#### List all knowledge bases +#### List all datasets ```python for ds in rag_object.list_datasets(): print(ds) ``` -#### Retrieve a knowledge base by ID +#### Retrieve a dataset by ID ```python dataset = rag_object.list_datasets(id = "id_1") @@ -191,23 +177,22 @@ print(dataset[0]) --- -## Update knowledge base +## Update dataset ```python DataSet.update(update_message: dict) ``` -Updates the current knowledge base. +Updates the current dataset. ### Parameters #### update_message: `dict[str, str|int]`, *Required* -- `"name"`: `str` The name of the knowledge base to update. -- `"tenant_id"`: `str` The `"tenant_id` you get after calling `create_dataset()`. ????????????????????? +- `"name"`: `str` The name of the dataset to update. - `"embedding_model"`: `str` The embedding model for generating vector embeddings. - Ensure that `"chunk_count"` is `0` before updating `"embedding_model"`. -- `"parser_method"`: `str` The default parsing method for the knowledge base. +- `"chunk_method"`: `str` The default parsing method for the dataset. - `"naive"`: General - `"manual`: Manual - `"qa"`: Q&A @@ -233,22 +218,24 @@ from ragflow import RAGFlow rag = RAGFlow(api_key="", base_url="http://:9380") dataset = rag.list_datasets(name="kb_name") -dataset.update({"embedding_model":"BAAI/bge-zh-v1.5", "parse_method":"manual"}) +dataset.update({"embedding_model":"BAAI/bge-zh-v1.5", "chunk_method":"manual"}) ``` --- :::tip API GROUPING -File Management within Knowledge Base +File Management within Dataset ::: +--- + ## Upload documents ```python DataSet.upload_documents(document_list: list[dict]) ``` -Updloads documents to the current knowledge base. +Uploads documents to the current dataset. ### Parameters @@ -256,9 +243,8 @@ Updloads documents to the current knowledge base. A list of dictionaries representing the documents to upload, each containing the following keys: -- `"name"`: (Optional) File path to the document to upload. - Ensure that each file path has a suffix. -- `"blob"`: (Optional) The document to upload in binary format. +- `"display_name"`: (Optional) The file name to display in the dataset. +- `"blob"`: (Optional) The binary content of the file to upload. ### Returns @@ -268,8 +254,8 @@ A list of dictionaries representing the documents to upload, each containing the ### Examples ```python -dataset = rag.create_dataset(name="kb_name") -dataset.upload_documents([{"name": "1.txt", "blob": "123"}]) +dataset = rag_object.create_dataset(name="kb_name") +dataset.upload_documents([{"display_name": "1.txt", "blob": ""}, {"display_name": "2.pdf", "blob": ""}]) ``` --- @@ -284,9 +270,27 @@ Updates configurations for the current document. ### Parameters -#### update_message: `dict[str, str|int]`, *Required* +#### update_message: `dict[str, str|dict[]]`, *Required* -only `name`, `parser_config`, and `parser_method` can be changed +- `"name"`: `str` The name of the document to update. +- `"parser_config"`: `dict[str, Any]` The parsing configuration for the document: + - `"chunk_token_count"`: Defaults to `128`. + - `"layout_recognize"`: Defaults to `True`. + - `"delimiter"`: Defaults to `'\n!?。；！？'`. + - `"task_page_size"`: Defaults to `12`. +- `"chunk_method"`: `str` The parsing method to apply to the document. + - `"naive"`: General + - `"manual`: Manual + - `"qa"`: Q&A + - `"table"`: Table + - `"paper"`: Paper + - `"book"`: Book + - `"laws"`: Laws + - `"presentation"`: Presentation + - `"picture"`: Picture + - `"one"`: One + - `"knowledge_graph"`: Knowledge Graph + - `"email"`: Email ### Returns @@ -303,7 +307,7 @@ dataset=rag.list_datasets(id='id') dataset=dataset[0] doc = dataset.list_documents(id="wdfxb5t547d") doc = doc[0] -doc.update([{"parser_method": "manual"}]) +doc.update([{"parser_config": {"chunk_token_count": 256}}, {"chunk_method": "manual"}]) ``` --- @@ -314,19 +318,21 @@ doc.update([{"parser_method": "manual"}]) Document.download() -> bytes ``` +Downloads the current document from RAGFlow. + ### Returns -Bytes of the document. +The downloaded document in bytes. ### Examples ```python from ragflow import RAGFlow -rag = RAGFlow(api_key="", base_url="http://:9380") -ds=rag.list_datasets(id="id") -ds=ds[0] -doc = ds.list_documents(id="wdfxb5t547d") +rag_object = RAGFlow(api_key="", base_url="http://:9380") +dataset = rag_object.list_datasets(id="id") +dataset = dataset[0] +doc = dataset.list_documents(id="wdfxb5t547d") doc = doc[0] open("~/ragflow.txt", "wb+").write(doc.download()) print(doc) @@ -340,15 +346,17 @@ print(doc) Dataset.list_documents(id:str =None, keywords: str=None, offset: int=0, limit:int = 1024,order_by:str = "create_time", desc: bool = True) -> list[Document] ``` +Retrieves a list of documents from the current dataset. + ### Parameters #### id -The id of the document to retrieve. +The ID of the document to retrieve. Defaults to `None`. #### keywords -List documents whose name has the given keywords. Defaults to `None`. +The keywords to match document titles. Defaults to `None`. #### offset @@ -360,11 +368,14 @@ Records number to return, `-1` means all of them. Records number to return, `-1` #### orderby -The field by which the records should be sorted. This specifies the attribute or column used to order the results. +The field by which the documents should be sorted. Available options: + +- `"create_time"` (Default) +- `"update_time"` #### desc -A boolean flag indicating whether the sorting should be in descending order. +Indicates whether the retrieved documents should be sorted in descending order. Defaults to `True`. ### Returns @@ -375,8 +386,8 @@ A `Document` object contains the following attributes: - `id` Id of the retrieved document. Defaults to `""`. - `thumbnail` Thumbnail image of the retrieved document. Defaults to `""`. -- `knowledgebase_id` Knowledge base ID related to the document. Defaults to `""`. -- `parser_method` Method used to parse the document. Defaults to `""`. +- `knowledgebase_id` Dataset ID related to the document. Defaults to `""`. +- `chunk_method` Method used to parse the document. Defaults to `""`. - `parser_config`: `ParserConfig` Configuration object for the parser. Defaults to `None`. - `source_type`: Source type of the document. Defaults to `""`. - `type`: Type or category of the document. Defaults to `""`. @@ -414,7 +425,7 @@ for d in dataset.list_documents(keywords="rag", offset=0, limit=12): DataSet.delete_documents(ids: list[str] = None) ``` -Deletes specified documents or all documents from the current knowledge base. +Deletes specified documents or all documents from the current dataset. ### Returns @@ -434,11 +445,10 @@ ds.delete_documents(ids=["id_1","id_2"]) --- -## Parse and stop parsing document +## Parse documents ```python DataSet.async_parse_documents(document_ids:list[str]) -> None -DataSet.async_cancel_parse_documents(document_ids:list[str])-> None ``` ### Parameters @@ -476,6 +486,47 @@ print("Async bulk parsing cancelled") --- +## Stop parsing documents + +```python +DataSet.async_cancel_parse_documents(document_ids:list[str])-> None +``` + +### Parameters + +#### document_ids: `list[str]` + +The IDs of the documents to stop parsing. + +### Returns + +- Success: No value is returned. +- Failure: `Exception` + +### Examples + +```python +#documents parse and cancel +rag = RAGFlow(API_KEY, HOST_ADDRESS) +ds = rag.create_dataset(name="dataset_name") +documents = [ + {'name': 'test1.txt', 'blob': open('./test_data/test1.txt',"rb").read()}, + {'name': 'test2.txt', 'blob': open('./test_data/test2.txt',"rb").read()}, + {'name': 'test3.txt', 'blob': open('./test_data/test3.txt',"rb").read()} +] +ds.upload_documents(documents) +documents=ds.list_documents(keywords="test") +ids=[] +for document in documents: + ids.append(document.id) +ds.async_parse_documents(ids) +print("Async bulk parsing initiated") +ds.async_cancel_parse_documents(ids) +print("Async bulk parsing cancelled") +``` + +--- + ## List chunks ```python @@ -590,13 +641,18 @@ doc.delete_chunks(["id_1","id_2"]) ```python Chunk.update(update_message: dict) ``` + +Updates the current chunk. + ### Parameters -#### update_message: *Required* +#### update_message: `dict[str, str|list[str]|int]` *Required* -- `content`: `str` Contains the main text or information of the chunk -- `important_keywords`: `list[str]` List the key terms or phrases that are significant or central to the chunk's content -- `available`: `int` Indicating the availability status, `0` means unavailable and `1` means available +- `"content"`: `str` Content of the chunk. +- `"important_keywords"`: `list[str]` A list of key terms to attach to the chunk. +- `"available"`: `int` The chunk's availability status in the dataset. + - `0`: Unavailable + - `1`: Available ### Returns @@ -608,8 +664,8 @@ Chunk.update(update_message: dict) ```python from ragflow import RAGFlow -rag = RAGFlow(api_key="", base_url="http://:9380") -dataset = rag.list_datasets(id="123") +rag_object = RAGFlow(api_key="", base_url="http://:9380") +dataset = rag_object.list_datasets(id="123") dataset = dataset[0] doc = dataset.list_documents(id="wdfxb5t547d") doc = doc[0] @@ -619,7 +675,7 @@ chunk.update({"content":"sdfx..."}) --- -## Retrieval +## Retrieve chunks ```python RAGFlow.retrieve(question:str="", datasets:list[str]=None, document=list[str]=None, offset:int=1, limit:int=30, similarity_threshold:float=0.2, vector_similarity_weight:float=0.3, top_k:int=1024,rerank_id:str=None,keyword:bool=False,higlight:bool=False) -> list[Chunk] @@ -627,25 +683,25 @@ RAGFlow.retrieve(question:str="", datasets:list[str]=None, document=list[str]=No ### Parameters -#### question: `str`, *Required* +#### question: `str` *Required* The user query or query keywords. Defaults to `""`. -#### datasets: `list[Dataset]`, *Required* +#### datasets: `list[str]`, *Required* -The scope of datasets. +The datasets to search from. -#### document: `list[Document]` +#### document: `list[str]` -The scope of document. `None` means no limitation. Defaults to `None`. +The documents to search from. `None` means no limitation. Defaults to `None`. #### offset: `int` -The beginning point of retrieved records. Defaults to `0`. +The beginning point of retrieved chunks. Defaults to `0`. #### limit: `int` -The maximum number of records needed to return. Defaults to `6`. +The maximum number of chunks to return. Defaults to `6`. #### Similarity_threshold: `float` @@ -653,40 +709,47 @@ The minimum similarity score. Defaults to `0.2`. #### similarity_threshold_weight: `float` -The weight of vector cosine similarity, 1 - x is the term similarity weight. Defaults to `0.3`. +The weight of vector cosine similarity. Defaults to `0.3`. If x represents the vector cosine similarity, then (1 - x) is the term similarity weight. #### top_k: `int` -Number of records engaged in vector cosine computaton. Defaults to `1024`. +The number of chunks engaged in vector cosine computaton. Defaults to `1024`. + +#### rerank_id -#### rerank_id:`str` -ID of the rerank model. Defaults to `None`. +The ID of the rerank model. Defaults to `None`. -#### keyword:`bool` -Indicating whether keyword-based matching is enabled (True) or disabled (False). +#### keyword + +Indicates whether keyword-based matching is enabled: + +- `True`: Enabled. +- `False`: Disabled. #### highlight:`bool` Specifying whether to enable highlighting of matched terms in the results (True) or not (False). + ### Returns -list[Chunk] +- Success: A list of `Chunk` objects representing the document chunks. +- Failure: `Exception` ### Examples ```python from ragflow import RAGFlow -rag = RAGFlow(api_key="", base_url="http://:9380") -ds = rag.list_datasets(name="ragflow") +rag_object = RAGFlow(api_key="", base_url="http://:9380") +ds = rag_object.list_datasets(name="ragflow") ds = ds[0] name = 'ragflow_test.txt' path = './test_data/ragflow_test.txt' -rag.create_document(ds, name=name, blob=open(path, "rb").read()) +rag_object.create_document(ds, name=name, blob=open(path, "rb").read()) doc = ds.list_documents(name=name) doc = doc[0] ds.async_parse_documents([doc.id]) -for c in rag.retrieve(question="What's ragflow?", +for c in rag_object.retrieve(question="What's ragflow?", datasets=[ds.id], documents=[doc.id], offset=1, limit=30, similarity_threshold=0.2, vector_similarity_weight=0.3, @@ -705,9 +768,9 @@ Chat Assistant Management ```python RAGFlow.create_chat( - name: str = "assistant", - avatar: str = "path", - knowledgebases: list[DataSet] = [], + name: str, + avatar: str = "", + knowledgebases: list[str] = [], llm: Chat.LLM = None, prompt: Chat.Prompt = None ) -> Chat @@ -715,51 +778,74 @@ RAGFlow.create_chat( Creates a chat assistant. -### Returns - -- Success: A `Chat` object representing the chat assistant. -- Failure: `Exception` +### Parameters The following shows the attributes of a `Chat` object: -- `name`: `str` The name of the chat assistant. Defaults to `"assistant"`. -- `avatar`: `str` Base64 encoding of the avatar. Defaults to `""`. -- `knowledgebases`: `list[str]` The associated knowledge bases. Defaults to `["kb1"]`. -- `llm`: `LLM` The llm of the created chat. Defaults to `None`. When the value is `None`, a dictionary with the following values will be generated as the default. - - `model_name`, `str` - The chat model name. If it is `None`, the user's default chat model will be returned. - - `temperature`, `float` - Controls the randomness of the model's predictions. A lower temperature increases the model's conficence in its responses; a higher temperature increases creativity and diversity. Defaults to `0.1`. - - `top_p`, `float` - Also known as “nucleus sampling”, this parameter sets a threshold to select a smaller set of words to sample from. It focuses on the most likely words, cutting off the less probable ones. Defaults to `0.3` - - `presence_penalty`, `float` - This discourages the model from repeating the same information by penalizing words that have already appeared in the conversation. Defaults to `0.2`. - - `frequency penalty`, `float` - Similar to the presence penalty, this reduces the model’s tendency to repeat the same words frequently. Defaults to `0.7`. - - `max_token`, `int` - This sets the maximum length of the model’s output, measured in the number of tokens (words or pieces of words). Defaults to `512`. -- `Prompt`: `Prompt` Instructions for the LLM to follow. - - `"similarity_threshold"`: `float` A similarity score to evaluate distance between two lines of text. It's weighted keywords similarity and vector cosine similarity. If the similarity between query and chunk is less than this threshold, the chunk will be filtered out. Defaults to `0.2`. - - `"keywords_similarity_weight"`: `float` It's weighted keywords similarity and vector cosine similarity or rerank score (0~1). Defaults to `0.7`. - - `"top_n"`: `int` Not all the chunks whose similarity score is above the 'similarity threshold' will be feed to LLMs. LLM can only see these 'Top N' chunks. Defaults to `8`. - - `"variables"`: `list[dict[]]` If you use dialog APIs, the variables might help you chat with your clients with different strategies. The variables are used to fill in the 'System' part in prompt in order to give LLM a hint. The 'knowledge' is a very special variable which will be filled-in with the retrieved chunks. All the variables in 'System' should be curly bracketed. Defaults to `[{"key": "knowledge", "optional": True}]` - - `"rerank_model"`: `str` If it is not specified, vector cosine similarity will be used; otherwise, reranking score will be used. Defaults to `""`. -- `"empty_response"`: `str` If nothing is retrieved in the knowledge base for the user's question, this will be used as the response. To allow the LLM to improvise when nothing is retrieved, leave this blank. Defaults to `None`. +#### name: *Required* + +The name of the chat assistant. Defaults to `"assistant"`. + +#### avatar + +Base64 encoding of the avatar. Defaults to `""`. + +#### knowledgebases: `list[str]` + +The IDs of the associated datasets. Defaults to `[""]`. + +#### llm + +The llm of the created chat. Defaults to `None`. When the value is `None`, a dictionary with the following values will be generated as the default. + +An `LLM` object contains the following attributes: + +- `model_name`, `str` + The chat model name. If it is `None`, the user's default chat model will be returned. +- `temperature`, `float` + Controls the randomness of the model's predictions. A lower temperature increases the model's conficence in its responses; a higher temperature increases creativity and diversity. Defaults to `0.1`. +- `top_p`, `float` + Also known as “nucleus sampling”, this parameter sets a threshold to select a smaller set of words to sample from. It focuses on the most likely words, cutting off the less probable ones. Defaults to `0.3` +- `presence_penalty`, `float` + This discourages the model from repeating the same information by penalizing words that have already appeared in the conversation. Defaults to `0.2`. +- `frequency penalty`, `float` + Similar to the presence penalty, this reduces the model’s tendency to repeat the same words frequently. Defaults to `0.7`. +- `max_token`, `int` + This sets the maximum length of the model’s output, measured in the number of tokens (words or pieces of words). Defaults to `512`. + +#### Prompt + +Instructions for the LLM to follow. A `Prompt` object contains the following attributes: + +- `"similarity_threshold"`: `float` A similarity score to evaluate distance between two lines of text. It's weighted keywords similarity and vector cosine similarity. If the similarity between query and chunk is less than this threshold, the chunk will be filtered out. Defaults to `0.2`. +- `"keywords_similarity_weight"`: `float` It's weighted keywords similarity and vector cosine similarity or rerank score (0~1). Defaults to `0.7`. +- `"top_n"`: `int` Not all the chunks whose similarity score is above the 'similarity threshold' will be feed to LLMs. LLM can only see these 'Top N' chunks. Defaults to `8`. +- `"variables"`: `list[dict[]]` If you use dialog APIs, the variables might help you chat with your clients with different strategies. The variables are used to fill in the 'System' part in prompt in order to give LLM a hint. The 'knowledge' is a very special variable which will be filled-in with the retrieved chunks. All the variables in 'System' should be curly bracketed. Defaults to `[{"key": "knowledge", "optional": True}]` +- `"rerank_model"`: `str` If it is not specified, vector cosine similarity will be used; otherwise, reranking score will be used. Defaults to `""`. +- `"empty_response"`: `str` If nothing is retrieved in the dataset for the user's question, this will be used as the response. To allow the LLM to improvise when nothing is retrieved, leave this blank. Defaults to `None`. - `"opener"`: `str` The opening greeting for the user. Defaults to `"Hi! I am your assistant, can I help you?"`. - `"show_quote`: `bool` Indicates whether the source of text should be displayed Defaults to `True`. -- `"prompt"`: `str` The prompt content. Defaults to `You are an intelligent assistant. Please summarize the content of the knowledge base to answer the question. Please list the data in the knowledge base and answer in detail. When all knowledge base content is irrelevant to the question, your answer must include the sentence "The answer you are looking for is not found in the knowledge base!" Answers need to consider chat history. +- `"prompt"`: `str` The prompt content. Defaults to `You are an intelligent assistant. Please summarize the content of the dataset to answer the question. Please list the data in the knowledge base and answer in detail. When all knowledge base content is irrelevant to the question, your answer must include the sentence "The answer you are looking for is not found in the knowledge base!" Answers need to consider chat history. Here is the knowledge base: {knowledge} The above is the knowledge base.`. +### Returns + +- Success: A `Chat` object representing the chat assistant. +- Failure: `Exception` + ### Examples ```python from ragflow import RAGFlow rag = RAGFlow(api_key="", base_url="http://:9380") -knowledge_base = rag.list_datasets(name="kb_1") -assistant = rag.create_chat("Miss R", knowledgebases=knowledge_base) +kbs = rag.list_datasets(name="kb_1") +list_kb=[] +for kb in kbs: + list_kb.append(kb.id) +assi = rag.create_chat("Miss R", knowledgebases=list_kb) ``` --- @@ -778,7 +864,7 @@ Updates the current chat assistant. - `"name"`: `str` The name of the chat assistant to update. - `"avatar"`: `str` Base64 encoding of the avatar. Defaults to `""` -- `"knowledgebases"`: `list[str]` Knowledge bases to update. +- `"knowledgebases"`: `list[str]` datasets to update. - `"llm"`: `dict` The LLM settings: - `"model_name"`, `str` The chat model name. - `"temperature"`, `float` Controls the randomness of the model's predictions. @@ -792,7 +878,7 @@ Updates the current chat assistant. - `"top_n"`: `int` Not all the chunks whose similarity score is above the 'similarity threshold' will be feed to LLMs. LLM can only see these 'Top N' chunks. Defaults to `8`. - `"variables"`: `list[dict[]]` If you use dialog APIs, the variables might help you chat with your clients with different strategies. The variables are used to fill in the 'System' part in prompt in order to give LLM a hint. The 'knowledge' is a very special variable which will be filled-in with the retrieved chunks. All the variables in 'System' should be curly bracketed. Defaults to `[{"key": "knowledge", "optional": True}]` - `"rerank_model"`: `str` If it is not specified, vector cosine similarity will be used; otherwise, reranking score will be used. Defaults to `""`. - - `"empty_response"`: `str` If nothing is retrieved in the knowledge base for the user's question, this will be used as the response. To allow the LLM to improvise when nothing is retrieved, leave this blank. Defaults to `None`. + - `"empty_response"`: `str` If nothing is retrieved in the dataset for the user's question, this will be used as the response. To allow the LLM to improvise when nothing is retrieved, leave this blank. Defaults to `None`. - `"opener"`: `str` The opening greeting for the user. Defaults to `"Hi! I am your assistant, can I help you?"`. - `"show_quote`: `bool` Indicates whether the source of text should be displayed Defaults to `True`. - `"prompt"`: `str` The prompt content. Defaults to `You are an intelligent assistant. Please summarize the content of the knowledge base to answer the question. Please list the data in the knowledge base and answer in detail. When all knowledge base content is irrelevant to the question, your answer must include the sentence "The answer you are looking for is not found in the knowledge base!" Answers need to consider chat history. @@ -879,7 +965,7 @@ The attribute by which the results are sorted. Defaults to `"create_time"`. #### desc -Indicates whether to sort the results in descending order. Defaults to `True`. +Indicates whether the retrieved chat assistants should be sorted in descending order. Defaults to `True`. #### id: `string` @@ -1017,25 +1103,25 @@ The content of the message. Defaults to `"Hi! I am your assistant, can I help yo A list of `Chunk` objects representing references to the message, each containing the following attributes: -- **id**: `str` +- `id` `str` The chunk ID. -- **content**: `str` +- `content` `str` The content of the chunk. -- **image_id**: `str` +- `image_id` `str` The ID of the snapshot of the chunk. -- **document_id**: `str` +- `document_id` `str` The ID of the referenced document. -- **document_name**: `str` +- `document_name` `str` The name of the referenced document. -- **position**: `list[str]` +- `position` `list[str]` The location information of the chunk within the referenced document. -- **knowledgebase_id**: `str` - The ID of the knowledge base to which the referenced document belongs. -- **similarity**: `float` +- `knowledgebase_id` `str` + The ID of the dataset to which the referenced document belongs. +- `similarity` `float` A composite similarity score of the chunk ranging from `0` to `1`, with a higher value indicating greater similarity. -- **vector_similarity**: `float` +- `vector_similarity` `float` A vector similarity score of the chunk ranging from `0` to `1`, with a higher value indicating greater similarity between vector embeddings. -- **term_similarity**: `float` +- `term_similarity` `float` A keyword similarity score of the chunk ranging from `0` to `1`, with a higher value indicating greater similarity between keywords. @@ -1091,11 +1177,14 @@ The number of records on each page. Defaults to `1024`. #### orderby -The field by which the records should be sorted. This specifies the attribute or column used to sort the results. Defaults to `"create_time"`. +The field by which the sessions should be sorted. Available options: + +- `"create_time"` (Default) +- `"update_time"` #### desc -Whether the sorting should be in descending order. Defaults to `True`. +Indicates whether the retrieved sessions should be sorted in descending order. Defaults to `True`. #### id