fireworks-mixtral test

gistfile1.txt

Here is the code and results of a test
<code>
#https://github.com/fw-ai/cookbook/blob/main/examples/rag/mongo_basic.ipynb
key="whatever"
model_name="accounts/fireworks/models/mixtral-8x7b-instruct"
prompt = """
The user message will contain a block of text drawn from a scientific paper. Please analyze this text and perform the following steps:

1. Extract Named Entities: Identify named entities in the text which could be of three types - Drug, Disease, or Other. An entity should be a single proper noun or a term that is clearly defined within the scope of Drugs and Diseases. Entity names should be short and focused, containing no more than 5 essential words. If an entity name cannot be expressed in 5 words or less, it should be ignored. 

2. Map Relationships: Establish relationships between these entities, based on their context in the text. Relationships should have short, descriptive names that do not include other nouns. If a relationship name cannot be expressed in 5 words or less, it should be ignored. 

3. Handle Abbreviations: If an entity name in the text has an abbreviation, treat the abbreviation as a distinct entity. The abbreviation should be linked to the full name via a "abbreviation of" relationship, and vice versa, the full name should have an "abbreviation" relationship with the abbreviation.

4. Format Findings: Organize your findings as a JSON object. In this object, each key should be a named entity, and its corresponding value should be another object. This nested object should describe the relationships of the entity (each relationship should be a separate key), the related entities (as an array of values if multiple entities are involved), and the entity type. 

Please include an additional key, "_ENTITY_TYPE", to classify the entity as either "Drug", "Disease", or "Other".

Here's an illustrative example. For a text like: "Tom Currier is a great guy who built lots of communities after he studied at Stanford University (SU) and Harvard. He also won the Thiel Fellowship.", the output could be:

{
"Tom Currier": {
  "studied at": ["Stanford University", "Harvard"],
  "winner of": "Thiel Fellowship"
  "_ENTITY_TYPE": "Other"
},
"Stanford University": {
  "students": ["Tom Currier"],
  "abbreviation": "SU"
  "_ENTITY_TYPE": "Other"
},
"SU": {
  "abbreviation of": "Stanford University",
  "_ENTITY_TYPE": "Other"
},
}

Responses must always be valid JSON objects. Make sure that all keys in both the top-level object and any nested objects have valid JSON values. 

If no entities or relationships are identifiable in the given text, respond with the phrase "NO_ENTITIES_FOUND". Ensure that your response is exclusively the JSON data of extracted entities and relationships, or the phrase "NO_ENTITIES_FOUND".
"""

text="""
NALP3, also known as cryopyrin, has been implicated in crystal-associated diseases such as gout and silicosis. Mice fed the diet high in soluble oxalate demonstrated increased NALP3 expression in the kidney.
"""
from fireworks.client import Fireworks

client = Fireworks(api_key=key)
response = client.chat.completions.create(
  model=model_name,
  max_tokens=2048,
  messages=[{
    "role": "system",
    "content": prompt},
    {"role": "user",
    "content": text}
])
print(response.choices[0].message.content)
</code>

Results
<json>
{
  "NALP3": {
    "_ENTITY_TYPE": "Drug"
  },
  "cryopyrin": {
    "abbreviation": "NALP3",
    "_ENTITY_TYPE": "Drug"
  },
  "gout": {
    "_ENTITY_TYPE": "Disease"
  },
  "silicosis": {
    "_ENTITY_TYPE": "Disease"
  },
  "mice": {
    "fed diet high in soluble oxalate": {
      "increased NALP3 expression": ["kidney"]
    },
    "_ENTITY_TYPE": "Other"
  },
  "diet high in soluble oxalate": {
    "increased NALP3 expression": ["kidney"],
    "_ENTITY_TYPE": "Other"
  },
  "kidney": {
    "increased NALP3 expression": ["diet high in soluble oxalate"],
    "_ENTITY_TYPE": "Other"
  }
}
</json>