AI/ragflow
1
0
Fork 0
mirror of https://git.mirrors.martin98.com/https://github.com/infiniflow/ragflow.git synced 2025-04-22 06:00:00 +08:00
Code Issues Packages Projects Releases Wiki Activity

DRAFT: Updated chunk APIs (#2901)

Browse Source 
### What problem does this PR solve?



### Type of change


- [x] Documentation Update

---------

Signed-off-by: Jin Hai <haijin.chn@gmail.com>
Co-authored-by: Jin Hai <haijin.chn@gmail.com>
This commit is contained in:
writinwaters 2024-10-18 20:56:33 +08:00 committed by GitHub
parent 526fcbbfde
commit cec208051f
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
2 changed files with 235 additions and 146 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
api/http_api.md
View File

@ -37,7 +37,7 @@ Creates a dataset.
# "name": name is required and can't be duplicated. # "name": name is required and can't be duplicated.
# "tenant_id": tenant_id must not be provided. # "tenant_id": tenant_id must not be provided.
# "embedding_model": embedding_model must not be provided. # "embedding_model": embedding_model must not be provided.
# "navie" means general. # "naive" means general.
curl --request POST \ curl --request POST \
--url http://{address}/api/v1/dataset \ --url http://{address}/api/v1/dataset \
--header 'Content-Type: application/json' \ --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. # "chunk_count": If you update chunk_count, it can't be changed.
# "document_count": If you update document_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. # "parse_method": If you update parse_method, chunk_count must be 0.
# "navie" means general. # "naive" means general.
curl --request PUT \ curl --request PUT \
--url http://{address}/api/v1/dataset/{dataset_id} \ --url http://{address}/api/v1/dataset/{dataset_id} \
--header 'Content-Type: application/json' \ --header 'Content-Type: application/json' \
@ -247,7 +247,7 @@ curl --request PUT \
"embedding_model": "BAAI/bge-zh-v1.5", "embedding_model": "BAAI/bge-zh-v1.5",
"chunk_count": 0, "chunk_count": 0,
"document_count": 0, "document_count": 0,
"parse_method": "navie" "parse_method": "naive"
}' }'
``` ```

375
api/python_api_reference.md
View File

@ -3,10 +3,10 @@
**THE API REFERENCES BELOW ARE STILL UNDER DEVELOPMENT.** **THE API REFERENCES BELOW ARE STILL UNDER DEVELOPMENT.**
:::tip NOTE :::tip NOTE
Knowledge Base Management Dataset Management
::: :::
## Create knowledge base ## Create dataset
```python ```python
RAGFlow.create_dataset( RAGFlow.create_dataset(
@ -17,12 +17,12 @@ RAGFlow.create_dataset(
permission: str = "me", permission: str = "me",
document_count: int = 0, document_count: int = 0,
chunk_count: int = 0, chunk_count: int = 0,
parse_method: str = "naive", chunk_method: str = "naive",
parser_config: DataSet.ParserConfig = None parser_config: DataSet.ParserConfig = None
) -> DataSet ) -> DataSet
``` ```
Creates a knowledge base (dataset). Creates a dataset.
### Parameters ### 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 `""` 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` #### description: `str`
The description of the created dataset. Defaults to `""`. A brief description of the dataset to create. Defaults to `""`.
#### language: `str` #### language: `str`
The language setting of the created dataset. Defaults to `"English"`. ???????????? The language setting of the dataset to create. Available options:
- `"English"` (Default)
- `"Chinese"`
#### permission #### permission
Specify who can operate on the dataset. Defaults to `"me"`. Specifies who can operate on the dataset. You can set it only to `"me"` for now.
#### document_count: `int` #### chunk_method, `str`
The number of documents associated with the dataset. Defaults to `0`. The default parsing method of the knwoledge . Defaults to `"naive"`.
#### chunk_count: `int`
The number of data chunks generated or processed by the created dataset. Defaults to `0`.
#### parse_method, `str`
The method used by the dataset to parse and process data. Defaults to `"naive"`.
#### parser_config #### parser_config
@ -100,19 +86,19 @@ ds = rag_object.create_dataset(name="kb_1")
--- ---
## Delete knowledge bases ## Delete datasets
```python ```python
RAGFlow.delete_datasets(ids: list[str] = None) RAGFlow.delete_datasets(ids: list[str] = None)
``` ```
Deletes knowledge bases by name or ID. Deletes datasets by name or ID.
### Parameters ### Parameters
#### ids #### ids
The IDs of the knowledge bases to delete. The IDs of the datasets to delete.
### Returns ### Returns
@ -127,7 +113,7 @@ rag.delete_datasets(ids=["id_1","id_2"])
--- ---
## List knowledge bases ## List datasets
```python ```python
RAGFlow.list_datasets( RAGFlow.list_datasets(
@ -140,7 +126,7 @@ RAGFlow.list_datasets(
) -> list[DataSet] ) -> list[DataSet]
``` ```
Retrieves a list of knowledge bases. Retrieves a list of datasets.
### Parameters ### Parameters
@ -158,7 +144,7 @@ The field by which the records should be sorted. This specifies the attribute or
#### desc: `bool` #### 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` #### id: `str`
@ -170,19 +156,19 @@ The name of the dataset to be got. Defaults to `None`.
### Returns ### Returns
- Success: A list of `DataSet` objects representing the retrieved knowledge bases. - Success: A list of `DataSet` objects representing the retrieved datasets.
- Failure: `Exception`. - Failure: `Exception`.
### Examples ### Examples
#### List all knowledge bases #### List all datasets
```python ```python
for ds in rag_object.list_datasets(): for ds in rag_object.list_datasets():
print(ds) print(ds)
``` ```
#### Retrieve a knowledge base by ID #### Retrieve a dataset by ID
```python ```python
dataset = rag_object.list_datasets(id = "id_1") dataset = rag_object.list_datasets(id = "id_1")
@ -191,23 +177,22 @@ print(dataset[0])
--- ---
## Update knowledge base ## Update dataset
```python ```python
DataSet.update(update_message: dict) DataSet.update(update_message: dict)
``` ```
Updates the current knowledge base. Updates the current dataset.
### Parameters ### Parameters
#### update_message: `dict[str, str|int]`, *Required* #### update_message: `dict[str, str|int]`, *Required*
- `"name"`: `str` The name of the knowledge base to update. - `"name"`: `str` The name of the dataset to update.
- `"tenant_id"`: `str` The `"tenant_id` you get after calling `create_dataset()`. ?????????????????????
- `"embedding_model"`: `str` The embedding model for generating vector embeddings. - `"embedding_model"`: `str` The embedding model for generating vector embeddings.
- Ensure that `"chunk_count"` is `0` before updating `"embedding_model"`. - 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 - `"naive"`: General
- `"manual`: Manual - `"manual`: Manual
- `"qa"`: Q&A - `"qa"`: Q&A
@ -233,22 +218,24 @@ from ragflow import RAGFlow
rag = RAGFlow(api_key="<YOUR_API_KEY>", base_url="http://<YOUR_BASE_URL>:9380") rag = RAGFlow(api_key="<YOUR_API_KEY>", base_url="http://<YOUR_BASE_URL>:9380")
dataset = rag.list_datasets(name="kb_name") 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 :::tip API GROUPING
File Management within Knowledge Base File Management within Dataset
::: :::
---
## Upload documents ## Upload documents
```python ```python
DataSet.upload_documents(document_list: list[dict]) DataSet.upload_documents(document_list: list[dict])
``` ```
Updloads documents to the current knowledge base. Uploads documents to the current dataset.
### Parameters ### 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: A list of dictionaries representing the documents to upload, each containing the following keys:
- `"name"`: (Optional) File path to the document to upload. - `"display_name"`: (Optional) The file name to display in the dataset.
Ensure that each file path has a suffix. - `"blob"`: (Optional) The binary content of the file to upload.
- `"blob"`: (Optional) The document to upload in binary format.
### Returns ### Returns
@ -268,8 +254,8 @@ A list of dictionaries representing the documents to upload, each containing the
### Examples ### Examples
```python ```python
dataset = rag.create_dataset(name="kb_name") dataset = rag_object.create_dataset(name="kb_name")
dataset.upload_documents([{"name": "1.txt", "blob": "123"}]) dataset.upload_documents([{"display_name": "1.txt", "blob": "<BINARY_CONTENT_OF_THE_DOC>"}, {"display_name": "2.pdf", "blob": "<BINARY_CONTENT_OF_THE_DOC>"}])
``` ```
--- ---
@ -284,9 +270,27 @@ Updates configurations for the current document.
### Parameters ### 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 ### Returns
@ -303,7 +307,7 @@ dataset=rag.list_datasets(id='id')
dataset=dataset[0] dataset=dataset[0]
doc = dataset.list_documents(id="wdfxb5t547d") doc = dataset.list_documents(id="wdfxb5t547d")
doc = doc[0] 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 Document.download() -> bytes
``` ```
Downloads the current document from RAGFlow.
### Returns ### Returns
Bytes of the document. The downloaded document in bytes.
### Examples ### Examples
```python ```python
from ragflow import RAGFlow from ragflow import RAGFlow
rag = RAGFlow(api_key="<YOUR_API_KEY>", base_url="http://<YOUR_BASE_URL>:9380") rag_object = RAGFlow(api_key="<YOUR_API_KEY>", base_url="http://<YOUR_BASE_URL>:9380")
ds=rag.list_datasets(id="id") dataset = rag_object.list_datasets(id="id")
ds=ds[0] dataset = dataset[0]
doc = ds.list_documents(id="wdfxb5t547d") doc = dataset.list_documents(id="wdfxb5t547d")
doc = doc[0] doc = doc[0]
open("~/ragflow.txt", "wb+").write(doc.download()) open("~/ragflow.txt", "wb+").write(doc.download())
print(doc) 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] 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 ### Parameters
#### id #### id
The id of the document to retrieve. The ID of the document to retrieve. Defaults to `None`.
#### keywords #### keywords
List documents whose name has the given keywords. Defaults to `None`. The keywords to match document titles. Defaults to `None`.
#### offset #### offset
@ -360,11 +368,14 @@ Records number to return, `-1` means all of them. Records number to return, `-1`
#### orderby #### 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 #### 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 ### Returns
@ -375,8 +386,8 @@ A `Document` object contains the following attributes:
- `id` Id of the retrieved document. Defaults to `""`. - `id` Id of the retrieved document. Defaults to `""`.
- `thumbnail` Thumbnail image 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 `""`. - `knowledgebase_id` Dataset ID related to the document. Defaults to `""`.
- `parser_method` Method used to parse 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`. - `parser_config`: `ParserConfig` Configuration object for the parser. Defaults to `None`.
- `source_type`: Source type of the document. Defaults to `""`. - `source_type`: Source type of the document. Defaults to `""`.
- `type`: Type or category 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) 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 ### Returns
@ -434,11 +445,10 @@ ds.delete_documents(ids=["id_1","id_2"])
--- ---
## Parse and stop parsing document ## Parse documents
```python ```python
DataSet.async_parse_documents(document_ids:list[str]) -> None DataSet.async_parse_documents(document_ids:list[str]) -> None
DataSet.async_cancel_parse_documents(document_ids:list[str])-> None
``` ```
### Parameters ### 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 ## List chunks
```python ```python
@ -590,13 +641,18 @@ doc.delete_chunks(["id_1","id_2"])
```python ```python
Chunk.update(update_message: dict) Chunk.update(update_message: dict)
``` ```
Updates the current chunk.
### Parameters ### Parameters
#### update_message: *Required* #### update_message: `dict[str, str|list[str]|int]` *Required*
- `content`: `str` Contains the main text or information of the chunk - `"content"`: `str` Content of the chunk.
- `important_keywords`: `list[str]` List the key terms or phrases that are significant or central to the chunk's content - `"important_keywords"`: `list[str]` A list of key terms to attach to the chunk.
- `available`: `int` Indicating the availability status, `0` means unavailable and `1` means available - `"available"`: `int` The chunk's availability status in the dataset.
- `0`: Unavailable
- `1`: Available
### Returns ### Returns
@ -608,8 +664,8 @@ Chunk.update(update_message: dict)
```python ```python
from ragflow import RAGFlow from ragflow import RAGFlow
rag = RAGFlow(api_key="<YOUR_API_KEY>", base_url="http://<YOUR_BASE_URL>:9380") rag_object = RAGFlow(api_key="<YOUR_API_KEY>", base_url="http://<YOUR_BASE_URL>:9380")
dataset = rag.list_datasets(id="123") dataset = rag_object.list_datasets(id="123")
dataset = dataset[0] dataset = dataset[0]
doc = dataset.list_documents(id="wdfxb5t547d") doc = dataset.list_documents(id="wdfxb5t547d")
doc = doc[0] doc = doc[0]
@ -619,7 +675,7 @@ chunk.update({"content":"sdfx..."})
--- ---
## Retrieval ## Retrieve chunks
```python ```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] 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 ### Parameters
#### question: `str`, *Required* #### question: `str` *Required*
The user query or query keywords. Defaults to `""`. 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` #### offset: `int`
The beginning point of retrieved records. Defaults to `0`. The beginning point of retrieved chunks. Defaults to `0`.
#### limit: `int` #### 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` #### Similarity_threshold: `float`
@ -653,40 +709,47 @@ The minimum similarity score. Defaults to `0.2`.
#### similarity_threshold_weight: `float` #### 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` #### 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:`str` #### rerank_id
ID of the rerank model. Defaults to `None`.
#### keyword:`bool` The ID of the rerank model. Defaults to `None`.
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` #### highlight:`bool`
Specifying whether to enable highlighting of matched terms in the results (True) or not (False). Specifying whether to enable highlighting of matched terms in the results (True) or not (False).
### Returns ### Returns
list[Chunk] - Success: A list of `Chunk` objects representing the document chunks.
- Failure: `Exception`
### Examples ### Examples
```python ```python
from ragflow import RAGFlow from ragflow import RAGFlow
rag = RAGFlow(api_key="<YOUR_API_KEY>", base_url="http://<YOUR_BASE_URL>:9380") rag_object = RAGFlow(api_key="<YOUR_API_KEY>", base_url="http://<YOUR_BASE_URL>:9380")
ds = rag.list_datasets(name="ragflow") ds = rag_object.list_datasets(name="ragflow")
ds = ds[0] ds = ds[0]
name = 'ragflow_test.txt' name = 'ragflow_test.txt'
path = './test_data/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 = ds.list_documents(name=name)
doc = doc[0] doc = doc[0]
ds.async_parse_documents([doc.id]) 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], datasets=[ds.id], documents=[doc.id],
offset=1, limit=30, similarity_threshold=0.2, offset=1, limit=30, similarity_threshold=0.2,
vector_similarity_weight=0.3, vector_similarity_weight=0.3,
@ -705,9 +768,9 @@ Chat Assistant Management
```python ```python
RAGFlow.create_chat( RAGFlow.create_chat(
name: str = "assistant", name: str,
avatar: str = "path", avatar: str = "",
knowledgebases: list[DataSet] = [], knowledgebases: list[str] = [],
llm: Chat.LLM = None, llm: Chat.LLM = None,
prompt: Chat.Prompt = None prompt: Chat.Prompt = None
) -> Chat ) -> Chat
@ -715,51 +778,74 @@ RAGFlow.create_chat(
Creates a chat assistant. Creates a chat assistant.
### Parameters
The following shows the attributes of a `Chat` object:
#### 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 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 ### Returns
- Success: A `Chat` object representing the chat assistant. - Success: A `Chat` object representing the chat assistant.
- Failure: `Exception` - Failure: `Exception`
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`.
- `"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.
Here is the knowledge base:
{knowledge}
The above is the knowledge base.`.
### Examples ### Examples
```python ```python
from ragflow import RAGFlow from ragflow import RAGFlow
rag = RAGFlow(api_key="<YOUR_API_KEY>", base_url="http://<YOUR_BASE_URL>:9380") rag = RAGFlow(api_key="<YOUR_API_KEY>", base_url="http://<YOUR_BASE_URL>:9380")
knowledge_base = rag.list_datasets(name="kb_1") kbs = rag.list_datasets(name="kb_1")
assistant = rag.create_chat("Miss R", knowledgebases=knowledge_base) 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. - `"name"`: `str` The name of the chat assistant to update.
- `"avatar"`: `str` Base64 encoding of the avatar. Defaults to `""` - `"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: - `"llm"`: `dict` The LLM settings:
- `"model_name"`, `str` The chat model name. - `"model_name"`, `str` The chat model name.
- `"temperature"`, `float` Controls the randomness of the model's predictions. - `"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`. - `"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}]` - `"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 `""`. - `"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?"`. - `"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`. - `"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 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 #### 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` #### 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: A list of `Chunk` objects representing references to the message, each containing the following attributes:
- **id**: `str` - `id` `str`
The chunk ID. The chunk ID.
- **content**: `str` - `content` `str`
The content of the chunk. The content of the chunk.
- **image_id**: `str` - `image_id` `str`
The ID of the snapshot of the chunk. The ID of the snapshot of the chunk.
- **document_id**: `str` - `document_id` `str`
The ID of the referenced document. The ID of the referenced document.
- **document_name**: `str` - `document_name` `str`
The name of the referenced document. The name of the referenced document.
- **position**: `list[str]` - `position` `list[str]`
The location information of the chunk within the referenced document. The location information of the chunk within the referenced document.
- **knowledgebase_id**: `str` - `knowledgebase_id` `str`
The ID of the knowledge base to which the referenced document belongs. The ID of the dataset to which the referenced document belongs.
- **similarity**: `float` - `similarity` `float`
A composite similarity score of the chunk ranging from `0` to `1`, with a higher value indicating greater similarity. 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. 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. 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 #### 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 #### 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 #### id