ton/SQLLLM
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

update

Browse Source
This commit is contained in:
narawat lamaiin
2024-10-14 09:12:06 +07:00
parent 7b2d85da48
commit 4ef968b86e
8 changed files with 473 additions and 432 deletions
Download Patch File Download Diff File Expand all files Collapse all files

36
src/interface.jl
View File

@@ -250,8 +250,7 @@ function decisionMaker(state::T1, context, text2textInstructLLM::Function,
error("DecisionMaker has more than one key per categories")
end
end
println("--> SQLLLM decisionMaker() ", @__FILE__, " ", @__LINE__)
pprintln(Dict(responsedict))
return responsedict
catch e
io = IOBuffer()
@@ -499,7 +498,7 @@ julia>
function evaluator(state::T1, text2textInstructLLM::Function;
addSQLVectorDB::Union{Function, Nothing}=nothing
) where {T1<:AbstractDict}
println("Evaluating state", @__FILE__, " ", @__LINE__)
# systemmsg =
# """
# You are a helpful assistant that analyzes agent's trajectories to find solutions and observations (i.e., the results of actions) to answer the user's questions.
@@ -737,14 +736,16 @@ function evaluator(state::T1, text2textInstructLLM::Function;
# mark as terminal state when the answer is achieved
if accepted_as_answer == "Yes"
state[:isterminal] = true
state[:reward] = 1
# user score as reward because different answers hold different value for the user.
state[:reward] = responsedict[:score]
#add to vectorDB
if addSQLVectorDB !== nothing
addSQLVectorDB(state)
end
end
println("--> 5 Evaluator ", @__FILE__, " ", @__LINE__)
println("~~~ 5 Evaluator() ", @__FILE__, " ", @__LINE__)
pprintln(Dict(responsedict))
return responsedict[:score]
@@ -953,7 +954,7 @@ julia> state = Dict(
# TODO
- [] add embedding of newstate and store in newstate[:embedding]
- [WORKING] should getdata() return isterminal?
# Signature
"""
function transition(state::T, args::NamedTuple
@@ -992,17 +993,13 @@ function transition(state::T, args::NamedTuple
else
error("undefined LLM function. Requesting $actionname")
end
# this section allow LLM functions above to have different return values.
result = haskey(response, :result) ? response[:result] : nothing
success::Bool = haskey(response, :success) ? response[:success] : false
result = success ? response[:result] : response[:errormsg]
select = haskey(response, :select) ? response[:select] : nothing
reward::Integer = haskey(response, :reward) ? response[:reward] : 0
isterminal::Bool = haskey(response, :isterminal) ? response[:isterminal] : false
errormsg::Union{AbstractString, Nothing} = haskey(response, :errormsg) ? response[:errormsg] : nothing
success::Bool = haskey(response, :success) ? response[:success] : false
newNodeKey, newstate = makeNewState(state, thoughtDict, JSON3.write(result), select, reward, isterminal)
println("SQLLLM transition() 1 ", @__FILE__, " ", @__LINE__)
progressvalue::Integer = evaluatorF(newstate, text2textInstructLLM;
addSQLVectorDB=addSQLVectorDBF)
@@ -1090,14 +1087,14 @@ function query(query::T, executeSQL::Function, text2textInstructLLM::Function;
addSQLVectorDB::Union{Function, Nothing}=nothing,
querySQLVectorDB::Union{Function, Nothing}=nothing
)::String where {T<:AbstractString}
#[WORKING] add extra context for Evaluator so that it knows the observation is from seaching a database
# add extra context for Evaluator so that it knows the observation is from seaching a database
query = "Search the database for {$query}"
initialstate = Dict{Symbol, Any}(
:reward=> 0,
:isterminal=> false,
:evaluation=> "None",
:suggestion=> "None",
:evaluationscore=> 0,
:suggestion=> "None",
:accepted_as_answer=> "No",
:lesson=> nothing,
@@ -1121,10 +1118,13 @@ function query(query::T, executeSQL::Function, text2textInstructLLM::Function;
addSQLVectorDB=addSQLVectorDB,
)
_, result = LLMMCTS.runMCTS(initialstate, transition, transitionargs;
totalsample=1, maxdepth=3, maxiterations=1, explorationweight=1.0)
latestKey, _ = GeneralUtils.findHighestIndexKey(result[:thoughtHistory], "observation")
resulttext = result[:thoughtHistory][latestKey]
earlystop(state) = state[:reward] >= 8 ? true : false
_, resultState = LLMMCTS.runMCTS(initialstate, transition, transitionargs;
totalsample=1, maxdepth=3, maxiterations=3, explorationweight=1.0,
earlystop=earlystop)
latestKey, _ = GeneralUtils.findHighestIndexKey(resultState[:thoughtHistory], "observation")
resulttext = resultState[:thoughtHistory][latestKey]
return resulttext
end

608
src/llmfunction.jl
View File

@@ -1,10 +1,10 @@
module llmfunction
export listAllTable_json, listAllTable_str, tableinfo, getdata, finalAnswerBox,
getTableNameFromSQL, extractContent_dataframe, SQLexecution
export listAllTable_json, listAllTable_str, tableinfo, getdata, finalAnswerBox,
getTableNameFromSQL, extractContent_dataframe, SQLexecution
using HTTP, JSON3, URIs, Random, PrettyPrinting, UUIDs, LibPQ, Tables, DataFrames, CSV,
DataStructures, StatsBase
using HTTP, JSON3, URIs, Random, PrettyPrinting, UUIDs, LibPQ, Tables, DataFrames, CSV,
DataStructures, StatsBase
using GeneralUtils, LLMMCTS
using ..util
@@ -36,25 +36,24 @@ julia> result = response[:result]
# Signature
"""
function listAllTable_json(executeSQL::Function
)::NamedTuple{(:result, :success), Tuple{DataFrame, Bool}}
)::NamedTuple{(:result, :success),Tuple{DataFrame,Bool}}
sql =
"""
SELECT
table_name,
obj_description(relfilenode, 'pg_class') AS table_comment,
string_agg(column_name || ' (' || data_type || ')', ', ') AS columns
FROM
information_schema.columns
JOIN
pg_class ON table_name = relname
WHERE
table_schema = 'public'
GROUP BY
table_name, relfilenode
ORDER BY
table_name;
"""
sql = """
SELECT
table_name,
obj_description(relfilenode, 'pg_class') AS table_comment,
string_agg(column_name || ' (' || data_type || ')', ', ') AS columns
FROM
information_schema.columns
JOIN
pg_class ON table_name = relname
WHERE
table_schema = 'public'
GROUP BY
table_name, relfilenode
ORDER BY
table_name;
"""
result = executeSQL(sql)
df = DataFrame(result)
@@ -65,24 +64,23 @@ end
function listAllTable_str(executeSQL::Function
)::NamedTuple{(:result, :success), Tuple{String, Bool}}
sql =
"""
SELECT
table_name,
obj_description(relfilenode, 'pg_class') AS table_comment,
string_agg(column_name || ' (' || data_type || ')', ', ') AS columns
FROM
information_schema.columns
JOIN
pg_class ON table_name = relname
WHERE
table_schema = 'public'
GROUP BY
table_name, relfilenode
ORDER BY
table_name;
"""
)::NamedTuple{(:result, :success),Tuple{String,Bool}}
sql = """
SELECT
table_name,
obj_description(relfilenode, 'pg_class') AS table_comment,
string_agg(column_name || ' (' || data_type || ')', ', ') AS columns
FROM
information_schema.columns
JOIN
pg_class ON table_name = relname
WHERE
table_schema = 'public'
GROUP BY
table_name, relfilenode
ORDER BY
table_name;
"""
result = executeSQL(sql)
df = DataFrame(result)
tableinfo = "Here are a list of available tables in the database (each row is in this format: table name; table comment; table columns): \n"
@@ -109,20 +107,19 @@ end
"""
function tableinfo_str(executeSQL::Function, tablename::String)::NamedTuple{(:result, :success), Tuple{String, Bool}}
sql =
"""
SELECT
column_name,
data_type,
col_description(format('%s.%s', table_schema, table_name)::regclass::oid, ordinal_position) AS column_comment
FROM
information_schema.columns
WHERE
table_name = '$tablename'
AND table_schema = 'public';
"""
function tableinfo_str(executeSQL::Function, tablename::String)::NamedTuple{(:result, :success),Tuple{String,Bool}}
sql = """
SELECT
column_name,
data_type,
col_description(format('%s.%s', table_schema, table_name)::regclass::oid, ordinal_position) AS column_comment
FROM
information_schema.columns
WHERE
table_name = '$tablename'
AND table_schema = 'public';
"""
result = executeSQL(sql)
df = DataFrame(result)
@@ -167,18 +164,17 @@ julia> result = response[:result]
# Signature
"""
function tableinfo(executeSQL::Function, tablenames::T
)::NamedTuple{(:result,), Tuple{String}} where {T<:AbstractVector}
)::NamedTuple{(:result,),Tuple{String}} where {T<:AbstractVector}
# list all tables in a database
sql =
"""
SELECT pg_namespace.nspname AS schema_name,
relname AS table_name,
pg_catalog.obj_description(pg_class.oid) AS comment
FROM pg_class
INNER JOIN pg_namespace ON pg_namespace.oid = pg_class.relnamespace
WHERE pg_namespace.nspname = 'public' -- Replace 'public' with your desired schema
AND pg_class.relkind IN ('r', 't');
"""
sql = """
SELECT pg_namespace.nspname AS schema_name,
relname AS table_name,
pg_catalog.obj_description(pg_class.oid) AS comment
FROM pg_class
INNER JOIN pg_namespace ON pg_namespace.oid = pg_class.relnamespace
WHERE pg_namespace.nspname = 'public' -- Replace 'public' with your desired schema
AND pg_class.relkind IN ('r', 't');
"""
_result = executeSQL(sql)
df = DataFrame(_result)
@@ -193,8 +189,7 @@ function tableinfo(executeSQL::Function, tablenames::T
end
end
if !isempty(notExistingTable)
result =
"Error, the following tables does not exist in the database: $(JSON3.write(notExistingTable))"
result = "Error, the following tables does not exist in the database: $(JSON3.write(notExistingTable))"
return (result=result,)
end
@@ -221,53 +216,70 @@ end
A connection object connected to the database
- `text2textInstructLLM::Function`
A function that handles communication to LLM service.
# Return
- `NamedTuple{(:result, :errormsg, success), Tuple{String, String, Bool}}`
# TODO
- [x] getdata directly using sql execute
# Signature
"""
function getdata(query::T, context::Union{Dict, Nothing}, executeSQL::Function,
function getdata(query::T, context::Union{Dict,Nothing}, executeSQL::Function,
text2textInstructLLM::Function;
)::NamedTuple{(:result, :errormsg, :success), Tuple{String, Union{String, Nothing}, Bool}} where {T<:AbstractString}
) where {T<:AbstractString}
# get table info here because it'll be called only 1-time. If this function is in
# getdata_decisionMaker(), it'll be called everytime
mentionedtable = getTableNameFromSQL(query, text2textInstructLLM)
mentionedTableInfo = tableinfo(executeSQL, mentionedtable)[:result]
context[:mentionedTableInfo] = mentionedTableInfo
initialstate = Dict{Symbol, Any}(
:reward=> 0,
:isterminal=> false,
:evaluation=> nothing,
:errormsg=> nothing,
:errorexplain=> nothing,
:question=> query,
:code=> nothing,
:response=> nothing,
)
transitionargs = (
# decisionMaker=getdata_decisionMaker,
# evaluator=getdata_evaluator,
# reflector=getdata_reflector,
context=context,
executeSQL=executeSQL,
text2textInstructLLM=text2textInstructLLM
)
result_1, result_2 = LLMMCTS.runMCTS(initialstate, getdata_transition, transitionargs;
totalsample=1, maxdepth=3, maxiterations=1, explorationweight=1.0)
if result_2[:isterminal] == true
return (result=result_2[:response], errormsg=nothing, success=true) # succues=true to finish getdata()
response = SQLexecution(executeSQL, query)
if response[:success]
extracted = extractContent_dataframe(response[:result], context, text2textInstructLLM)
response_ = (result=extracted, errormsg=nothing, success=true)
return response_
else
# return (response="Failed to act with the following error message: $(result_2[:errorexplain])", select=nothing, reward=0, success=false)
return (result="Failed to get the data. $(result_1[:errormsg])",
errormsg=result_1[:errormsg], success=false)
response_ = (result=nothing, errormsg=response[:errormsg], success=false)
return response_
end
end
# function getdata(query::T, context::Union{Dict, Nothing}, executeSQL::Function,
# text2textInstructLLM::Function;
# )::NamedTuple{(:result, :errormsg, :success), Tuple{String, Union{String, Nothing}, Bool}} where {T<:AbstractString}
# # get table info here because it'll be called only 1-time. If this function is in
# # getdata_decisionMaker(), it'll be called everytime
# mentionedtable = getTableNameFromSQL(query, text2textInstructLLM)
# mentionedTableInfo = tableinfo(executeSQL, mentionedtable)[:result]
# context[:mentionedTableInfo] = mentionedTableInfo
# initialstate = Dict{Symbol, Any}(
# :reward=> 0,
# :isterminal=> false,
# :evaluation=> nothing,
# :errormsg=> nothing,
# :errorexplain=> nothing,
# :question=> query,
# :code=> nothing,
# :response=> nothing,
# )
# transitionargs = (
# # decisionMaker=getdata_decisionMaker,
# # evaluator=getdata_evaluator,
# # reflector=getdata_reflector,
# context=context,
# executeSQL=executeSQL,
# text2textInstructLLM=text2textInstructLLM
# )
# result_1, result_2 = LLMMCTS.runMCTS(initialstate, getdata_transition, transitionargs;
# totalsample=1, maxdepth=3, maxiterations=1, explorationweight=1.0)
# if result_2[:isterminal] == true
# return (result=result_2[:response], errormsg=nothing, success=true) # succues=true to finish getdata()
# else
# # return (response="Failed to act with the following error message: $(result_2[:errorexplain])", select=nothing, reward=0, success=false)
# return (result="Failed to get the data. $(result_1[:errormsg])",
# errormsg=result_1[:errormsg], success=false)
# end
# end
"""
@@ -290,7 +302,7 @@ julia>
# Signature
"""
function getdata_evaluator(newstate, config)
return (evaluation="None", score=0)
end
@@ -309,9 +321,9 @@ end
# Signature
"""
function getdata_transition(state::T, args::NamedTuple
)::NamedTuple{(:newNodeKey, :newstate, :progressvalue), Tuple{String, T, Integer}} where {T<:AbstractDict}
)::NamedTuple{(:newNodeKey, :newstate, :progressvalue),Tuple{String,T,Integer}} where {T<:AbstractDict}
# decisionMaker::Function = args[:decisionMaker]
# evaluator::Function = args[:evaluator]
# reflector::Function = args[:reflector]
@@ -320,26 +332,26 @@ function getdata_transition(state::T, args::NamedTuple
text2textInstructLLM::Function = args[:text2textInstructLLM]
thought, sql =
if state[:code] !== nothing
result = getdata_decisionMaker(state, context, text2textInstructLLM)
result[:thought], result[:code]
else
nothing, state[:question]
end
if state[:code] !== nothing
result = getdata_decisionMaker(state, context, text2textInstructLLM)
result[:thought], result[:code]
else
nothing, state[:question]
end
# make new state
newNodeKey = GeneralUtils.uuid4snakecase()
newstate = deepcopy(state)
response, success, errormsg, reward, isterminal =
if sql !== nothing
response, success, errormsg, reward, isterminal = SQLexecution(executeSQL, sql)
else
(result= nothing,
success= false,
errormsg= "SQL execution failed. An unexpected error occurred. Please try again.",
reward=0,
isterminal=false)
(result=nothing,
success=false,
errormsg="SQL execution failed. An unexpected error occurred. Please try again.",
reward=0,
isterminal=false)
end
println("getdata_transition() 1 ", @__FILE__, " ", @__LINE__)
newstate[:code] = sql
@@ -376,10 +388,10 @@ end
# Signature
"""
function getdata_decisionMaker(state::Dict, context::Dict, text2textInstructLLM::Function
)::NamedTuple{(:thought, :code, :success, :errormsg), Tuple{Union{String, Nothing}, Union{String, Nothing}, Bool, Union{String, Nothing}}}
)::NamedTuple{(:thought, :code, :success, :errormsg),Tuple{Union{String,Nothing},Union{String,Nothing},Bool,Union{String,Nothing}}}
Hints = "None"
# """
# Here are some useful SQL programs:
# $usefulSQL
@@ -414,78 +426,75 @@ function getdata_decisionMaker(state::Dict, context::Dict, text2textInstructLLM:
# 2) ...
# ...
# code: ...
# Let's begin!
# """
systemmsg =
"""
You are an assistant helping the user to execute SQL code from the user's query.
systemmsg = """
You are an assistant helping the user to execute SQL code from the user's query.
At each round of conversation, the user will give you:
Context: ...
User intention: ...
Code executed from the last round: ...
Execution error: execution error of the last round code.
At each round of conversation, the user will give you:
Context: ...
User intention: ...
Code executed from the last round: ...
Execution error: execution error of the last round code.
You should consider the following guidelines:
- Text information in the database is sometimes stored in lower case. If your search returns empty, try using lower case to search.
You should consider the following guidelines:
- Text information in the database is sometimes stored in lower case. If your search returns empty, try using lower case to search.
You should then respond to the user with:
1) Understanding:
- State your understanding about the current situation.
2) Reasoning:
- State your step by step reasoning about the current situation.
3) Plan: Step-by-step instructions of how to complete the task.
- Focus on improving the code from the last round.
- Do not create any table in the database.
4) Code:
- Write new improved code.
- Do not wrap the code and no comment as it will be executed directly without any modification against the database.
You should then respond to the user with:
1) Understanding:
- State your understanding about the current situation.
2) Reasoning:
- State your step by step reasoning about the current situation.
3) Plan: Step-by-step instructions of how to complete the task.
- Focus on improving the code from the last round.
- Do not create any table in the database.
4) Code:
- Write new improved code.
- Do not wrap the code and no comment as it will be executed directly without any modification against the database.
You should only respond in format as described below and nothing more:
Understanding: ...
Reasoning: ...
Plan:
1) ...
2) ...
...
Code: ...
Let's begin!
"""
You should only respond in format as described below and nothing more:
Understanding: ...
Reasoning: ...
Plan:
1) ...
2) ...
...
Code: ...
Let's begin!
"""
noise = ""
note_flag = ""
for attempt in 1:10
usermsg =
"""
Context:
$(context[:mentionedTableInfo])
User intention: $(context[:userintention])
Code executed from the last round: $(state[:code])
Execution error: $(state[:errormsg])
$noise
$note_flag
"""
_prompt =
[
Dict(:name=> "system", :text=> systemmsg),
Dict(:name=> "user", :text=> usermsg)
]
usermsg = """
Context:
$(context[:mentionedTableInfo])
User intention: $(context[:userintention])
Code executed from the last round: $(state[:code])
Execution error: $(state[:errormsg])
$noise
$note_flag
"""
_prompt =
[
Dict(:name => "system", :text => systemmsg),
Dict(:name => "user", :text => usermsg)
]
# put in model format
prompt = GeneralUtils.formatLLMtext(_prompt; formatname="llama3instruct")
prompt *=
"""
<|start_header_id|>assistant<|end_header_id|>
"""
prompt *= """
<|start_header_id|>assistant<|end_header_id|>
"""
try
response = text2textInstructLLM(prompt)
responsedict = GeneralUtils.textToDict(response,
["Understanding", "Reasoning", "Plan", "Code"];
rightmarker=":", symbolkey=true, lowercasekey=true)
["Understanding", "Reasoning", "Plan", "Code"];
rightmarker=":", symbolkey=true, lowercasekey=true)
_code = responsedict[:code]
code = strip(_code)
@@ -509,7 +518,7 @@ function getdata_decisionMaker(state::Dict, context::Dict, text2textInstructLLM:
else
end
println("--> getdata_decisionMaker() ", @__FILE__, " ", @__LINE__)
println("~~~ getdata_decisionMaker() ", @__FILE__, " ", @__LINE__)
pprintln(Dict(responsedict))
return (thought=responsedict[:reasoning], code=code, success=true, errormsg=nothing)
catch e
@@ -523,7 +532,7 @@ function getdata_decisionMaker(state::Dict, context::Dict, text2textInstructLLM:
end
end
return (thought=nothing, code=nothing, success=false,
errormsg="Failed to generate SQL after numerous attempts.")
errormsg="Failed to generate SQL after numerous attempts.")
end
""" Execute a given SQL.
@@ -553,7 +562,7 @@ julia> response = SQLLLM.SQLexecution(executeSQL, sql)
"""
# function SQLexecution(executeSQL::Function, sql::T
# )::NamedTuple{(:result, :success, :errormsg, :reward, :isterminal), Tuple{Union{DataFrame, Nothing}, Bool, Union{String, Nothing}, Integer, Bool}} where {T<:AbstractString}
# println("--> 1-01 ", @__FILE__, " ", @__LINE__)
# println("~~~ 1-01 ", @__FILE__, " ", @__LINE__)
# #XXX dummy SQL. use for testing
# # sql = "SELECT w.wine_name FROM wine w JOIN wine_food wf ON w.wine_id = wf.wine_id JOIN food f ON wf.food_id = f.food_id WHERE f.\"food_name\" = 'lamb';"
# # sql = " SELECT w.wine_name FROM wine w JOIN food f ON f.food_name = 'lamb' JOIN wine_food wf ON w.wine_id = wf.wine_id AND f.food_id = wf.food_id GROUP BY w.wine_name ORDER BY COUNT(DISTINCT w.wine_id) DESC;"
@@ -567,37 +576,37 @@ julia> response = SQLLLM.SQLexecution(executeSQL, sql)
# # add LIMIT to the SQL to prevent loading large data
# sql = strip(sql)
# println("--> SQL 1", @__FILE__, " ", @__LINE__)
# println("~~~ SQL 1", @__FILE__, " ", @__LINE__)
# println(sql)
# println("--> 1-02 ", @__FILE__, " ", @__LINE__)
# println("~~~ 1-02 ", @__FILE__, " ", @__LINE__)
# if sql[end] != ';'
# errorMsg = "Error, SQL execution failed because it does not ended with ';'"
# return (result=nothing, success=false, errormsg=errorMsg, reward=0, isterminal=false)
# end
# println("--> 1-03 ", @__FILE__, " ", @__LINE__)
# println("~~~ 1-03 ", @__FILE__, " ", @__LINE__)
# if !occursin("LIMIT", sql)
# # sql = sql[1:end-1] * " LIMIT 100;"
# sql = sql[1:end-1] * " ORDER BY RANDOM() LIMIT 2;"
# end
# println("--> SQL 2", @__FILE__, " ", @__LINE__)
# println("~~~ SQL 2", @__FILE__, " ", @__LINE__)
# println(sql)
# println("--> 1-1 ", @__FILE__, " ", @__LINE__)
# println("~~~ 1-1 ", @__FILE__, " ", @__LINE__)
# result = executeSQL(sql)
# println("--> 1-2 ", @__FILE__, " ", @__LINE__)
# println("~~~ 1-2 ", @__FILE__, " ", @__LINE__)
# df = DataFrame(result)
# println("--> raw df ", df)
# println("~~~ raw df ", df)
# tablesize = size(df)
# println("--> df size ", tablesize)
# println("--> 6 ", @__FILE__, " ", @__LINE__)
# println("~~~ df size ", tablesize)
# println("~~~ 6 ", @__FILE__, " ", @__LINE__)
# row = tablesize[1]
# println("--> 7 ", @__FILE__, " ", @__LINE__)
# println("~~~ 7 ", @__FILE__, " ", @__LINE__)
# if row == 0 # if 0 row
# errorMsg = "The resulting table has 0 row. Possible causes: 1) SQL is incorrect 2) There is no data that match your search criteria."
# return (result=nothing, success=false, errormsg=errorMsg, reward=0, isterminal=false)
# end
# println("--> 8 ", @__FILE__, " ", @__LINE__)
# println("~~~ 8 ", @__FILE__, " ", @__LINE__)
# df1 =
# if row > 2
# # ramdom row to pick
@@ -606,13 +615,13 @@ julia> response = SQLLLM.SQLexecution(executeSQL, sql)
# df
# end
# println("--> SQLexecution result ", @__FILE__, " ", @__LINE__)
# println("~~~ SQLexecution result ", @__FILE__, " ", @__LINE__)
# println(df1)
# return (result=df1, success=true, errormsg=nothing, reward=1, isterminal=true)
# end
function SQLexecution(executeSQL::Function, sql::T
)::NamedTuple{(:result, :success, :errormsg, :reward, :isterminal), Tuple{Union{DataFrame, Nothing}, Bool, Union{String, Nothing}, Integer, Bool}} where {T<:AbstractString}
) where {T<:AbstractString}
try
#XXX dummy SQL. use for testing
# sql = "SELECT w.wine_name FROM wine w JOIN wine_food wf ON w.wine_id = wf.wine_id JOIN food f ON wf.food_id = f.food_id WHERE f.\"food_name\" = 'lamb';"
@@ -635,7 +644,7 @@ function SQLexecution(executeSQL::Function, sql::T
else
error("Error, SQL execution failed because it does not ended with ';'")
end
println("--> SQL ", @__FILE__, " ", @__LINE__)
println("~~~ SQL ", @__FILE__, " ", @__LINE__)
println(sql)
result = executeSQL(sql)
@@ -649,24 +658,28 @@ function SQLexecution(executeSQL::Function, sql::T
error("SQL execution failed. An unexpected error occurred. Please try again.")
end
df1 =
if row > 2
# ramdom row to pick
df[sample(1:nrow(df), 2, replace=false), :] # random select 2 rows from df
else
df
end
df1 =
if row > 2
# ramdom row to pick
df[sample(1:nrow(df), 2, replace=false), :] # random select 2 rows from df
else
df
end
println("--> SQLexecution() ", @__FILE__, " ", @__LINE__)
println("~~~ SQLexecution() ", @__FILE__, " ", @__LINE__)
println(df1)
return (result=df1, success=true, errormsg=nothing, reward=1, isterminal=true)
return (result=df1, success=true, errormsg=nothing)
catch e
println("~~~ Error SQLexecution() 2 ", @__FILE__, " ", @__LINE__)
io = IOBuffer()
showerror(io, e)
errorMsg = String(take!(io))
st = sprint((io, v) -> show(io, "text/plain", v), stacktrace(catch_backtrace()))
println(errorMsg)
return (result=nothing, success=false, errormsg=errorMsg, reward=0, isterminal=false)
println("~~~ Error SQLexecution() 2.1 ", @__FILE__, " ", @__LINE__)
response = (result=nothing, success=false, errormsg=errorMsg)
println("~~~ Error SQLexecution() 2.2 ", @__FILE__, " ", @__LINE__)
return response
end
end
@@ -687,100 +700,90 @@ end
# Signature
"""
function extractContent_dataframe(df::DataFrame, context::Dict, text2textInstructLLM::Function
)::String
)::String
tablesize = size(df)
row = tablesize[1]
column = tablesize[2]
#[PENDING] Since selected column depend on the question, there should be a better way to select column on the fly, not hard coded like this.
df1 =
if column > 10 # assuming if columns > 10, agent is getting wine info but the info is too much
selectedcolumn = ["wine_id",
"wine_name",
"brand",
"manufacturer",
"region",
"country",
"wine_type",
"grape_variety",
"serving_temperature",
"intensity",
"sweetness",
"tannin",
"acidity",
"fizziness",
"tasting_notes"]
df1 = df[:, selectedcolumn]
else
df
end
# df1 =
# if column > 10 # assuming if columns > 10, agent is getting wine info but the info is too much
# selectedcolumn = ["wine_id",
# "wine_name",
# "winery",
# "region",
# "country",
# "wine_type",
# "grape",
# "serving_temperature",
# "intensity",
# "sweetness",
# "tannin",
# "acidity",
# "fizziness",
# "tasting_notes"]
# df1 = df[:, selectedcolumn]
# else
# df
# end
df1 = df
dfstr = dfToString(df1)
# println("--> df string")
# println(dfstr)
systemmsg =
"""
You are an assistant that readouts the resulting table after the user executing SQL command.
systemmsg = """
You are an assistant that readouts the resulting table after the user executing SQL command.
At each round of conversation, the user will give you:
- User intention: ...
- Resulting table dimension: ...
- Resulting table: The resulting table after executing the user's intention.
At each round of conversation, the user will give you:
- User intention: ...
- Resulting table dimension: ...
- Resulting table: The resulting table after executing the user's intention.
You should then respond to the user with:
- About_resulting_table:
1) What is the resulting table represent?
- Search_summary:
1) Summarize the table's content based on the user intension in verbal English.
Here are some example:
Bad example (you are not Summarize the table content): there are 2 columns in the table i.e. "cash" and "number".
2) Do not generate additional text.
You should then respond to the user with:
- About_resulting_table:
1) What is the resulting table represent?
- Search_summary:
1) Summarize the table's content based on the user intension in verbal English.
Here are some example:
Bad example (you are not Summarize the table content): there are 2 columns in the table i.e. "cash" and "number".
2) Do not generate additional text.
You should only respond in format as described below:
About_resulting_table: ...
Search_summary: ...
Let's begin!
"""
You should only respond in format as described below:
About_resulting_table: ...
Search_summary: ...
usermsg =
"""
User intention: $(context[:userintention])
Resulting table: $dfstr
"""
_prompt =
[
Dict(:name=> "system", :text=> systemmsg),
Dict(:name=> "user", :text=> usermsg)
]
Let's begin!
"""
usermsg = """
User intention: $(context[:userintention])
Resulting table: $dfstr
"""
_prompt =
[
Dict(:name => "system", :text => systemmsg),
Dict(:name => "user", :text => usermsg)
]
# put in model format
prompt = GeneralUtils.formatLLMtext(_prompt; formatname="llama3instruct")
prompt *=
"""
<|start_header_id|>assistant<|end_header_id|>
"""
prompt *= """
<|start_header_id|>assistant<|end_header_id|>
"""
for i in 1:5
response = text2textInstructLLM(prompt)
responsedict = GeneralUtils.textToDict(response, ["About_resulting_table", "Search_summary"],
rightmarker=":", symbolkey=true)
responsedict = GeneralUtils.textToDict(response, ["About_resulting_table", "Search_summary"],
rightmarker=":", symbolkey=true)
# result = dfstr
result =
"""
Summary: $(responsedict[:Search_summary])
More details: $dfstr
"""
result = """
Summary: $(responsedict[:Search_summary])
More details: $dfstr
"""
if row > 2
result *= "There are many more rows, but they are truncated because there are too many of them."
end
println("--> extractContent_dataframe() ", @__FILE__, " ", @__LINE__)
println("~~~ extractContent_dataframe() ", @__FILE__, " ", @__LINE__)
println(result)
return result
@@ -856,47 +859,44 @@ julia> result = SQLLLM.getTableNameFromSQL(sql, text2textInstructLLM)
# Signature
"""
function getTableNameFromSQL(sql::T, text2textInstructLLM::Function)::Vector{String} where {T<:AbstractString}
systemmsg =
"""
Extract table name out of the user query.
systemmsg = """
Extract table name out of the user query.
At each round of conversation, the user will give you:
Query: ...
At each round of conversation, the user will give you:
Query: ...
You should then respond to the user with:
- table_name: a list of table name that the user mentioned in the query.
For example, ["color", "type"]
You should then respond to the user with:
- table_name: a list of table name that the user mentioned in the query.
For example, ["color", "type"]
You must only respond in format as described below:
table_name: ["...", "...", ...]
Let's begin!
"""
You must only respond in format as described below:
table_name: ["...", "...", ...]
usermsg =
"""
Query: $sql
"""
_prompt =
[
Dict(:name=> "system", :text=> systemmsg),
Dict(:name=> "user", :text=> usermsg)
]
Let's begin!
"""
usermsg = """
Query: $sql
"""
_prompt =
[
Dict(:name => "system", :text => systemmsg),
Dict(:name => "user", :text => usermsg)
]
# put in model format
prompt = GeneralUtils.formatLLMtext(_prompt; formatname="llama3instruct")
prompt *=
"""
<|start_header_id|>assistant<|end_header_id|>
"""
prompt *= """
<|start_header_id|>assistant<|end_header_id|>
"""
for attempt in 1:5
try
response = text2textInstructLLM(prompt)
responsedict = GeneralUtils.textToDict(response,
["table_name"],
rightmarker=":", symbolkey=true)
["table_name"],
rightmarker=":", symbolkey=true)
response = copy(JSON3.read(responsedict[:table_name]))
return response