Icon

Parsing Raw Data with LLMs

LLMs are good at parsing raw data into structured data.

If you don't want to understand the intricacies of raw data, usually scraped data, you can simply ask an LLM to do it for you. Simply provide the required JSON schema with the prompt.

URL: LLMs are very good at parsing - Hacker News https://news.ycombinator.com/item?id=34885601

As an LLM agent, your primary role is to parse this html text in a well structured JSON. Answer only with the JSON output and nothing else.

Here is the relevant context:

[RAW TEXT HERE]

Export the list of reviews context using the following JSON schema:

{'Review': {'properties': {'content': {'description': 'The content of the review.', 'title': 'Content', 'type': 'string'}, 'author': {'description': 'The name of the author.', 'title': 'Author', 'type': 'string'}, 'author_link': {'description': 'The link to the author profile.', 'title': 'Author Link', 'type': 'string'}, 'role': {'description': 'The role of the author.', 'title': 'Role', 'type': 'string'}, 'profile_image_link': {'description': 'The URL to the image of the author.', 'title': 'Profile Image Link', 'type': 'string'}}, 'required': ['content', 'author', 'author_link', 'role', 'profile_image_link'], 'title': 'Review', 'type': 'object'}}, 'properties': {'review': {'items': {'$ref': '#/$defs/Review'}, 'title': 'Review', 'type': 'array'}}

OpenAI Authenticator
4k tokensrequested at the output (the max)the chunking by 12 + prompting is roughly 2k tokens at the input (when it fails it tells you)8k is this GPT context length so we fit plenty
OpenAI Chat Model Connector
Chat Model Prompter
prompts initialization
Table Creator
Column Appender
join
String Manipulation
desired JSON schema
Table Creator
Column Appender
extracted data
Table View
Credentials Configuration
html from LinkedInRecommendations
File Reader (Complex Format)
concatenate cells
GroupBy
format instructions
Table Creator
replace
String Manipulation
divide using<li> tag
Cell Splitter
Table Transposer
chunking to fit context length
Chunk Loop Start
concatenate cells
GroupBy
empty conversation
Table Creator
Table Row to Variable
String to JSON
Row Filter
Column Filter
collect JSONs
Loop End
Row Filter
JSON schema validation
JSON Path

Nodes

Extensions

Links