Transform
Summarize
Summarize a large block of text into a smaller set of extracted sentences.
Parameters
-
length: The desired summary length in characters
-
match: Text to prioritize building the summary around
Example Usage 1
I have 20 cats and 40 dogs - it's a lot of furry friends to take care of!
My name is Jane and I run an animal rescue shelter out of my home.
It all started a few years ago when I took in a litter of abandoned kittens.
I fell in love with them and decided to make it my mission to give unwanted animals a forever home.
{{ summarize|length:100 }}
Example Usage 2 - Using match
I have 20 cats and 40 dogs - it's a lot of furry friends to take care of!
My name is Jane and I run an animal rescue shelter out of my home.
It all started a few years ago when I took in a litter of abandoned kittens.
I fell in love with them and decided to make it my mission to give unwanted animals a forever home.
{{ summarize|length:100|match:"love" }}
Translate
Translate input text.
Parameters
-
target: The 2-char language code
-
additionalInstructions: Optional - Additional instructions to the LLM
Example Usage
Would yield:Truncate by Tokens
Truncate text to a max number of LLM tokens.
Parameters
- tokens: The number of tokens to truncate to. A token is typically 3-4 characters.
Example Usage
{{ kb | query: "NeuralSeek" }}=>{{ truncateToken | tokens: "2000" }}=>{{ variable | name: "documentation" }}
documentation
variable. This helps prevent exceeding context windows of some smaller LLMs.
Remove Stopwords
Extract stopwords (either our defaults or the ones you have set on the Configure tab) from input text.
Example Usage
Will yield Notice the wordsI, have, and, isn't, this
are deemed as stopwords and thus have been removed.
Force Numeric
Extract numbers from text. When getting numerical imput from users, LLM's or external sources always use the forceNumeric node to cleanse them for downstream use. If multiple numbers are found in the input they will return in an array.
Example Usage
I have 20 cats and 40 dogs
contains numeric values. So, running this:
2040
Table Prep
Take CSV input data and reduce it down to the rows and columns that most likely contain the answer to a query, for use in later steps or to send to an LLM.
Parameters
-
query: The Question to ask of the table
-
sentences: Return as sentences, an array, or an object
Example Usage 1
If we have CSV data, table prep will convert it to JSON or natural language:
The result will be: