vortosql/config.yaml at main · momalekpour/vortosql · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
# NL2SQL pipeline configuration
nl2sql_pipeline:
  intent_guardrail:
    chat_completion_model_provider: openai      # LLM provider: "openai"
    chat_completion_model_name: gpt-5.4         # Model name for the provider above: "gpt-5.4"
    temperature: 0                              # 0 = deterministic; range 0–2

  schema_linker:
    db_file_path: data/employees.db             # Path to the SQLite database file
    technique: full                             # "full" = pass entire schema; "tcsl"/"scsl" = LLM-driven selection
    model_provider: openai                      # Required only for tcsl/scsl techniques
    model_name: gpt-5.4-mini                    # Required only for tcsl/scsl techniques (mini to reduce cost)
    temperature: 0                              # 0 = deterministic

  example_selector:
    technique: question_similarity              # "random" = random sample; "question_similarity" = embedding-based
    number_of_examples: 3                       # Number of few-shot examples to retrieve
    embedding_model_provider: openai            # Required for question_similarity technique
    embedding_model_name: text-embedding-3-small  # Embedding model to use
    random_seed: null                           # Integer for reproducibility, or null

  sql_generator:
    prompt_template: zero_shot                  # "zero_shot" = no examples; "few_shot" = use example_selector output
    chat_completion_model_provider: openai      # LLM provider
    chat_completion_model_name: gpt-5.4         # Model name
    temperature: 0                              # 0 = deterministic
    random_seed: null                           # Integer for reproducibility, or null

  sql_corrector:
    max_correction_attempts: 2                 # Number of retry attempts; 0 = disabled
    prompt_template: syntax_correction         # Correction prompt style; currently only "syntax_correction"
    dbms: sqlite                               # Target DBMS for SQL dialect: "sqlite" or "duckdb"
    chat_completion_model_provider: openai     # LLM provider
    chat_completion_model_name: gpt-5.4        # Model name
    temperature: 0                             # 0 = deterministic
    random_seed: null                          # Integer for reproducibility, or null

  sql_executor:
    db_file_path: data/employees.db            # Path to the SQLite database file
    dbms: sqlite                               # Target DBMS: "sqlite" or "duckdb"

  answer_generator:
    chat_completion_model_provider: openai      # LLM provider
    chat_completion_model_name: gpt-5.4-mini    # Model name (mini is sufficient for summarization)
    temperature: 0                              # 0 = deterministic