langchain/chains/graph_qa/prompts.py

# flake8: noqa
from langchain_core.prompts.prompt import PromptTemplate

_DEFAULT_ENTITY_EXTRACTION_TEMPLATE = """Extract all entities from the following text. As a guideline, a proper noun is generally capitalized. You should definitely extract all names and places.

Return the output as a single comma-separated list, or NONE if there is nothing of note to return.

EXAMPLE
i'm trying to improve Langchain's interfaces, the UX, its integrations with various products the user might want ... a lot of stuff.
Output: Langchain
END OF EXAMPLE

EXAMPLE
i'm trying to improve Langchain's interfaces, the UX, its integrations with various products the user might want ... a lot of stuff. I'm working with Sam.
Output: Langchain, Sam
END OF EXAMPLE

Begin!

{input}
Output:"""
ENTITY_EXTRACTION_PROMPT = PromptTemplate(
    input_variables=["input"], template=_DEFAULT_ENTITY_EXTRACTION_TEMPLATE
)

_DEFAULT_GRAPH_QA_TEMPLATE = """Use the following knowledge triplets to answer the question at the end. If you don't know the answer, just say that you don't know, don't try to make up an answer.

{context}

Question: {question}
Helpful Answer:"""
GRAPH_QA_PROMPT = PromptTemplate(
    template=_DEFAULT_GRAPH_QA_TEMPLATE, input_variables=["context", "question"]
)

CYPHER_GENERATION_TEMPLATE = """Task:Generate Cypher statement to query a graph database.
Instructions:
Use only the provided relationship types and properties in the schema.
Do not use any other relationship types or properties that are not provided.
Schema:
{schema}
Note: Do not include any explanations or apologies in your responses.
Do not respond to any questions that might ask anything else than for you to construct a Cypher statement.
Do not include any text except the generated Cypher statement.

The question is:
{question}"""
CYPHER_GENERATION_PROMPT = PromptTemplate(
    input_variables=["schema", "question"], template=CYPHER_GENERATION_TEMPLATE
)

NEBULAGRAPH_EXTRA_INSTRUCTIONS = """
Instructions:

First, generate cypher then convert it to NebulaGraph Cypher dialect(rather than standard):
1. it requires explicit label specification only when referring to node properties: v.`Foo`.name
2. note explicit label specification is not needed for edge properties, so it's e.name instead of e.`Bar`.name
3. it uses double equals sign for comparison: `==` rather than `=`
For instance:
```diff
< MATCH (p:person)-[e:directed]->(m:movie) WHERE m.name = 'The Godfather II'
< RETURN p.name, e.year, m.name;
---
> MATCH (p:`person`)-[e:directed]->(m:`movie`) WHERE m.`movie`.`name` == 'The Godfather II'
> RETURN p.`person`.`name`, e.year, m.`movie`.`name`;
```\n"""

NGQL_GENERATION_TEMPLATE = CYPHER_GENERATION_TEMPLATE.replace(
    "Generate Cypher", "Generate NebulaGraph Cypher"
).replace("Instructions:", NEBULAGRAPH_EXTRA_INSTRUCTIONS)

NGQL_GENERATION_PROMPT = PromptTemplate(
    input_variables=["schema", "question"], template=NGQL_GENERATION_TEMPLATE
)

KUZU_EXTRA_INSTRUCTIONS = """
Instructions:

Generate statement with Kùzu Cypher dialect (rather than standard):
1. do not use `WHERE EXISTS` clause to check the existence of a property because Kùzu database has a fixed schema.
2. do not omit relationship pattern. Always use `()-[]->()` instead of `()->()`.
3. do not include any notes or comments even if the statement does not produce the expected result.
```\n"""

KUZU_GENERATION_TEMPLATE = CYPHER_GENERATION_TEMPLATE.replace(
    "Generate Cypher", "Generate Kùzu Cypher"
).replace("Instructions:", KUZU_EXTRA_INSTRUCTIONS)

KUZU_GENERATION_PROMPT = PromptTemplate(
    input_variables=["schema", "question"], template=KUZU_GENERATION_TEMPLATE
)

GREMLIN_GENERATION_TEMPLATE = CYPHER_GENERATION_TEMPLATE.replace("Cypher", "Gremlin")

GREMLIN_GENERATION_PROMPT = PromptTemplate(
    input_variables=["schema", "question"], template=GREMLIN_GENERATION_TEMPLATE
)

CYPHER_QA_TEMPLATE = """You are an assistant that helps to form nice and human understandable answers.
The information part contains the provided information that you must use to construct an answer.
The provided information is authoritative, you must never doubt it or try to use your internal knowledge to correct it.
Make the answer sound as a response to the question. Do not mention that you based the result on the given information.
If the provided information is empty, say that you don't know the answer.
Information:
{context}

Question: {question}
Helpful Answer:"""
CYPHER_QA_PROMPT = PromptTemplate(
    input_variables=["context", "question"], template=CYPHER_QA_TEMPLATE
)

SPARQL_INTENT_TEMPLATE = """Task: Identify the intent of a prompt and return the appropriate SPARQL query type.
You are an assistant that distinguishes different types of prompts and returns the corresponding SPARQL query types.
Consider only the following query types:
* SELECT: this query type corresponds to questions
* UPDATE: this query type corresponds to all requests for deleting, inserting, or changing triples
Note: Be as concise as possible.
Do not include any explanations or apologies in your responses.
Do not respond to any questions that ask for anything else than for you to identify a SPARQL query type.
Do not include any unnecessary whitespaces or any text except the query type, i.e., either return 'SELECT' or 'UPDATE'.

The prompt is:
{prompt}
Helpful Answer:"""
SPARQL_INTENT_PROMPT = PromptTemplate(
    input_variables=["prompt"], template=SPARQL_INTENT_TEMPLATE
)

SPARQL_GENERATION_SELECT_TEMPLATE = """Task: Generate a SPARQL SELECT statement for querying a graph database.
For instance, to find all email addresses of John Doe, the following query in backticks would be suitable:
```
PREFIX foaf: <http://xmlns.com/foaf/0.1/>
SELECT ?email
WHERE {{
    ?person foaf:name "John Doe" .
    ?person foaf:mbox ?email .
}}
```
Instructions:
Use only the node types and properties provided in the schema.
Do not use any node types and properties that are not explicitly provided.
Include all necessary prefixes.
Schema:
{schema}
Note: Be as concise as possible.
Do not include any explanations or apologies in your responses.
Do not respond to any questions that ask for anything else than for you to construct a SPARQL query.
Do not include any text except the SPARQL query generated.

The question is:
{prompt}"""
SPARQL_GENERATION_SELECT_PROMPT = PromptTemplate(
    input_variables=["schema", "prompt"], template=SPARQL_GENERATION_SELECT_TEMPLATE
)

SPARQL_GENERATION_UPDATE_TEMPLATE = """Task: Generate a SPARQL UPDATE statement for updating a graph database.
For instance, to add 'jane.doe@foo.bar' as a new email address for Jane Doe, the following query in backticks would be suitable:
```
PREFIX foaf: <http://xmlns.com/foaf/0.1/>
INSERT {{
    ?person foaf:mbox <mailto:jane.doe@foo.bar> .
}}
WHERE {{
    ?person foaf:name "Jane Doe" .
}}
```
Instructions:
Make the query as short as possible and avoid adding unnecessary triples.
Use only the node types and properties provided in the schema.
Do not use any node types and properties that are not explicitly provided.
Include all necessary prefixes.
Schema:
{schema}
Note: Be as concise as possible.
Do not include any explanations or apologies in your responses.
Do not respond to any questions that ask for anything else than for you to construct a SPARQL query.
Return only the generated SPARQL query, nothing else.

The information to be inserted is:
{prompt}"""
SPARQL_GENERATION_UPDATE_PROMPT = PromptTemplate(
    input_variables=["schema", "prompt"], template=SPARQL_GENERATION_UPDATE_TEMPLATE
)

SPARQL_QA_TEMPLATE = """Task: Generate a natural language response from the results of a SPARQL query.
You are an assistant that creates well-written and human understandable answers.
The information part contains the information provided, which you can use to construct an answer.
The information provided is authoritative, you must never doubt it or try to use your internal knowledge to correct it.
Make your response sound like the information is coming from an AI assistant, but don't add any information.
Information:
{context}

Question: {prompt}
Helpful Answer:"""
SPARQL_QA_PROMPT = PromptTemplate(
    input_variables=["context", "prompt"], template=SPARQL_QA_TEMPLATE
)


AQL_GENERATION_TEMPLATE = """Task: Generate an ArangoDB Query Language (AQL) query from a User Input.

You are an ArangoDB Query Language (AQL) expert responsible for translating a `User Input` into an ArangoDB Query Language (AQL) query.

You are given an `ArangoDB Schema`. It is a JSON Object containing:
1. `Graph Schema`: Lists all Graphs within the ArangoDB Database Instance, along with their Edge Relationships.
2. `Collection Schema`: Lists all Collections within the ArangoDB Database Instance, along with their document/edge properties and a document/edge example.

You may also be given a set of `AQL Query Examples` to help you create the `AQL Query`. If provided, the `AQL Query Examples` should be used as a reference, similar to how `ArangoDB Schema` should be used.

Things you should do:
- Think step by step.
- Rely on `ArangoDB Schema` and `AQL Query Examples` (if provided) to generate the query.
- Begin the `AQL Query` by the `WITH` AQL keyword to specify all of the ArangoDB Collections required.
- Return the `AQL Query` wrapped in 3 backticks (```).
- Use only the provided relationship types and properties in the `ArangoDB Schema` and any `AQL Query Examples` queries.
- Only answer to requests related to generating an AQL Query.
- If a request is unrelated to generating AQL Query, say that you cannot help the user.

Things you should not do:
- Do not use any properties/relationships that can't be inferred from the `ArangoDB Schema` or the `AQL Query Examples`. 
- Do not include any text except the generated AQL Query.
- Do not provide explanations or apologies in your responses.
- Do not generate an AQL Query that removes or deletes any data.

Under no circumstance should you generate an AQL Query that deletes any data whatsoever.

ArangoDB Schema:
{adb_schema}

AQL Query Examples (Optional):
{aql_examples}

User Input:
{user_input}

AQL Query: 
"""

AQL_GENERATION_PROMPT = PromptTemplate(
    input_variables=["adb_schema", "aql_examples", "user_input"],
    template=AQL_GENERATION_TEMPLATE,
)

AQL_FIX_TEMPLATE = """Task: Address the ArangoDB Query Language (AQL) error message of an ArangoDB Query Language query.

You are an ArangoDB Query Language (AQL) expert responsible for correcting the provided `AQL Query` based on the provided `AQL Error`. 

The `AQL Error` explains why the `AQL Query` could not be executed in the database.
The `AQL Error` may also contain the position of the error relative to the total number of lines of the `AQL Query`.
For example, 'error X at position 2:5' denotes that the error X occurs on line 2, column 5 of the `AQL Query`.  

You are also given the `ArangoDB Schema`. It is a JSON Object containing:
1. `Graph Schema`: Lists all Graphs within the ArangoDB Database Instance, along with their Edge Relationships.
2. `Collection Schema`: Lists all Collections within the ArangoDB Database Instance, along with their document/edge properties and a document/edge example.

You will output the `Corrected AQL Query` wrapped in 3 backticks (```). Do not include any text except the Corrected AQL Query.

Remember to think step by step.

ArangoDB Schema:
{adb_schema}

AQL Query:
{aql_query}

AQL Error:
{aql_error}

Corrected AQL Query:
"""

AQL_FIX_PROMPT = PromptTemplate(
    input_variables=[
        "adb_schema",
        "aql_query",
        "aql_error",
    ],
    template=AQL_FIX_TEMPLATE,
)

AQL_QA_TEMPLATE = """Task: Generate a natural language `Summary` from the results of an ArangoDB Query Language query.

You are an ArangoDB Query Language (AQL) expert responsible for creating a well-written `Summary` from the `User Input` and associated `AQL Result`.

A user has executed an ArangoDB Query Language query, which has returned the AQL Result in JSON format.
You are responsible for creating an `Summary` based on the AQL Result.

You are given the following information:
- `ArangoDB Schema`: contains a schema representation of the user's ArangoDB Database.
- `User Input`: the original question/request of the user, which has been translated into an AQL Query.
- `AQL Query`: the AQL equivalent of the `User Input`, translated by another AI Model. Should you deem it to be incorrect, suggest a different AQL Query.
- `AQL Result`: the JSON output returned by executing the `AQL Query` within the ArangoDB Database.

Remember to think step by step.

Your `Summary` should sound like it is a response to the `User Input`.
Your `Summary` should not include any mention of the `AQL Query` or the `AQL Result`.

ArangoDB Schema:
{adb_schema}

User Input:
{user_input}

AQL Query:
{aql_query}

AQL Result:
{aql_result}
"""
AQL_QA_PROMPT = PromptTemplate(
    input_variables=["adb_schema", "user_input", "aql_query", "aql_result"],
    template=AQL_QA_TEMPLATE,
)


NEPTUNE_OPENCYPHER_EXTRA_INSTRUCTIONS = """
Instructions:
Generate the query in openCypher format and follow these rules:
Do not use `NONE`, `ALL` or `ANY` predicate functions, rather use list comprehensions.
Do not use `REDUCE` function. Rather use a combination of list comprehension and the `UNWIND` clause to achieve similar results.
Do not use `FOREACH` clause. Rather use a combination of `WITH` and `UNWIND` clauses to achieve similar results.{extra_instructions}
\n"""

NEPTUNE_OPENCYPHER_GENERATION_TEMPLATE = CYPHER_GENERATION_TEMPLATE.replace(
    "Instructions:", NEPTUNE_OPENCYPHER_EXTRA_INSTRUCTIONS
)

NEPTUNE_OPENCYPHER_GENERATION_PROMPT = PromptTemplate(
    input_variables=["schema", "question", "extra_instructions"],
    template=NEPTUNE_OPENCYPHER_GENERATION_TEMPLATE,
)

NEPTUNE_OPENCYPHER_GENERATION_SIMPLE_TEMPLATE = """
Write an openCypher query to answer the following question. Do not explain the answer. Only return the query.{extra_instructions}
Question:  "{question}". 
Here is the property graph schema: 
{schema}
\n"""

NEPTUNE_OPENCYPHER_GENERATION_SIMPLE_PROMPT = PromptTemplate(
    input_variables=["schema", "question", "extra_instructions"],
    template=NEPTUNE_OPENCYPHER_GENERATION_SIMPLE_TEMPLATE,
)