Documentation

Initialise pipeline

Mandatory first step to initialise a pipeline.

• CreateExtractionRequest

  Returns a request id and authorisation token needed to apply all other functions to the same batch of data. Sample Python code

Load data

Provide contextual problem domain data used to create knowledge graphs and, optionally, Case data.

• AddExtractionFile

  Add a single file (.csv, .pdf, .txt, or .json format).

• AddExtractionFiles

  Add multiple files (.csv, .pdf, .txt, or .json format).

• AddExtractionFileStream

  Add a data stream from a single file (.csv, .pdf, .txt, or .json format).

• AddExtractionFileStreams

  Add a data stream from multiple files (.csv, .pdf, .txt, or .json format).

Process data

Creates and stores documents and vector embeddings.

• ProcessExtract

  Extract text data from all (previously added) files. Data is then processed and stored: chunk text, create vector embeddings and metadata. Support for JSON-based extraction. CSV files, headers and corresponding values are converted to a map "[{'key': 'value', ...},{}...]"

• ProcessExtractFile

  Extract text data from specific (previously added) file(s). Data is then processed and stored: chunk text, create vector embeddings and metadata. Support for JSON-based extraction. CSV files, headers and corresponding values are converted to a map "[{'key': 'value', ...},{}...]"

• GetExtract

  Return processed data.

• GetExtractDocs

  Return processed data as Langchain documents.

Knowledge layer

Creates and stores knowledge graph elements.

• GetGraphStructure

  Extract named entities, relationships and contextual properties using NLP and LLM-based methods. Output graph elements as a JSON file in the format: [{"term": "Term/Node/Vertex name", "element": "Associative Term/Node/Vertex name", "relationships": "Relationship between Term/Node/Vertex names"},.....{}]

• GetGraphStructurePerDoc

  Per specified source document - extract named entities, relationships and contextual properties using NLP and LLM-based methods. Output graph elements as a JSON file in the format: [{"term": "Term/Node/Vertex name", "element": "Associative Term/Node/Vertex name", "relationships": "Relationship between Term/Node/Vertex names"},.....{}]

• GenerateGraph

  Write graph data to a graph database by submitting the JSON file generated by GetGraphStructure. Process completes and returns 'casesAdded n, associationsAdded n, documentsAdded n'.

• GenerateGraphPerDoc

  Per specified source document - write graph data to a graph database by submitting the JSON file generated by GetGraphStructure. Process completes and returns 'casesAdded n, associationsAdded n, documentsAdded n'.

• GenerateAutoGraph

  Please note, this function combines API calls from GetGraphStructurePerDoc and GenerateGraphPerDoc and is not featured in the API documentation. Accept a list of file names (refFiles) in place of refDoc and fileObj, that will be processed per file to create a full graph.

Retrieval

Graph based retrieval augmented generation (GraphRAG).

• GetGraph

  Return the full graph in JSON format.

• GetGraphRetrieval

  Submit and return an LLM query generated by the graph structure and content. With options to i) limit the search and retrieval to a single source document, and ii) select a base LLM model (open-source options only).

• GetGraphRetrievalWithDS

  Submit and return an LLM chat query generated by the graph structure and content with all relevant graph elements in JSON format. With options to i) limit the search and retrieval to a single source document, and ii) select a base LLM model (open-source options only).

• GetCustomerGraphRetrievalWithDS

  Accepts additional parameters for client projects managed by Tikos. Note that this function utilises the GetGraphRetrievalWithDS API call and includes project-specific parameters that are not documented publicly.

Transparency & Explainability

Integration with ML/AI models and final transparency and explainability functions.

• UploadModel

  Upload trained deep neural network model that needs to be embedded with Tikos Reasoning Platform. PyTorch based models are supported

• UploadModelConfig

  Upload the configuration related to the uploaded deep neural network model. Will accept the model param definition in JSON format as well as the model specification in YAML format

• UploadModelCaseData

  Upload of the selected knowledge Cases (feature sets), that will build the initial Sequential Collection Case base

• ProcessModel

  Process the uploaded deep neural network model with Synapses Logger embedding and dynamically creating the Sequential Collection Case base

• BuildSequentialCollection

  Write Sequential Collection (SC) knowledge structures for each Case and store them as new nodes and node properties. Sequential Collection is a proprietary technology used in the Tikos Reasoning Platform.

• GetSimilarCase

  Providing a Problem Space (PS) case, the Sequential Collection will conduct a basic binary (BIN, default) search or advanced binary (BINADV) search and return the most similar existing case. This does not perform any case adaptation.

• GetReasoning

  Performs reasoning from the problem case to proposed solution case candidates, undertakes case adaptation, returns the final solution, and saves the solution case to the case base.

Tikos Client

A generic client connector that orchestrates commonly used base functions. It has been developed to facilitate easy integration with other applications and supports multithreading.

• addProcessFiles

  Multithreading supported file processing function. Accepts a list of filenames and file paths as a tuple.

• addFileStreams

  Multithreading supported file addition function. Accepts a list of filenames and file stream as a tuple.

• addProcessFileStreams

  Multithreading supported combined file addition and processing function. Accepts a list of filenames and file stream as a tuple.

• generateGraphStructures

  Multithreading supported graph structure generation function. Accepts a list of filenames as contexts.

• createGraph

  Multithreading supported graph creation function. Accepts a list of filenames as context.

• getGraph

  Graph structure extraction function.

• getGraphRetrieval

  Graph retrieval function. Accepts filenames as context and query.

• uploadEmbeddingModel

  Supports upload of the deep neural network model

• uploadEmbeddingConfig

  Supports upload of the deep neural network model configuration files

• uploadModelCaseData

  Upload of the selected Knowledge Cases (feature sets), that will build the initial Sequential Collection Case base

• processEmbeddedModel

  Process the upload deep neural network model with Synapses Logger embedding and dynamically creating the Sequential Collection Case base

• createSequentialCollection

  Sequential Collection creation function. Accepts Case-Type, Data File name as context and Weight Type.

• generateReasoning

  Sequential Collection reasoning function. Accepts Case-Type, Data File name as context, problem space case as a JSON object string, Weight Type and Reasoning Type.