qfa.settings#

Module attributes

`LLM_RETRY_BUDGET_MULTIPLIER`	Worst-case retry budget of a single `LLMPort.complete` call, expressed as a multiple of the per-attempt `timeout`.
`DEFAULT_EMBEDDING_BATCH_SIZE`	Default records-per-onnxruntime-batch for the embedder.

Classes

`AnalyzeSettings`(_case_sensitive, ...)	Configuration specific to the `POST /v1/analyze` endpoint.
`AppSettings`(_case_sensitive, ...)	Root configuration composing all sub-settings groups.
`AuthSettings`([_case_sensitive, ...])	Configuration for API-key based authentication.
`DatabaseSettings`([_case_sensitive, ...])	Configuration for the PostgreSQL database connection.
`EmbeddingSettings`([_case_sensitive, ...])	Configuration for the self-hosted embedding model.
`JudgeLLMSettings`([_case_sensitive, ...])	Optional per-field overrides for the LLM connection used by judge calls.
`LLMSettings`([_case_sensitive, ...])	Configuration for the LLM provider connection.
`LogSettings`([_case_sensitive, ...])	Define settings for the logger.
`NetworkSettings`([_case_sensitive, ...])	Configuration for network settings.
`OrchestratorSettings`(_case_sensitive, ...)	Cross-cutting configuration shared by every orchestrator use case.
`TelemetrySettings`([_case_sensitive, ...])	Azure Monitor / Application Insights telemetry configuration.

qfa.settings.LLM_RETRY_BUDGET_MULTIPLIER: float = 3.0#: Worst-case retry budget of a single LLMPort.complete call, expressed as a multiple of the per-attempt timeout. The LLM adapter retries transient failures (timeout, rate-limit) up to LLM_RETRY_BUDGET_MULTIPLIER * timeout of wall-clock; LLMCallExecutor divides the deadline-derived budget by the same factor when sizing a per-attempt timeout, so even the worst-case retry sequence of the last call in a phase still finishes before the request deadline. The adapter and the executor MUST read this one constant so the two stay in lock-step — they live in different layers and cannot share code.

qfa.settings.DEFAULT_EMBEDDING_BATCH_SIZE = 100#

Default records-per-onnxruntime-batch for the embedder.

Single source of truth shared by EmbeddingSettings.batch_size and the qfa.adapters.embedding constructor/factory defaults, so the configurable default and the library default cannot silently drift apart.

class qfa.settings.LogSettings(_case_sensitive: bool | None = None, _nested_model_default_partial_update: bool | None = None, _env_prefix: str | None = None, _env_prefix_target: EnvPrefixTarget | None = None, _env_file: DotenvType | None = PosixPath('.'), _env_file_encoding: str | None = None, _env_ignore_empty: bool | None = None, _env_nested_delimiter: str | None = None, _env_nested_max_split: int | None = None, _env_parse_none_str: str | None = None, _env_parse_enums: bool | None = None, _cli_prog_name: str | None = None, _cli_parse_args: bool | list[str] | tuple[str, ...] | None = None, _cli_settings_source: CliSettingsSource[Any] | None = None, _cli_parse_none_str: str | None = None, _cli_hide_none_type: bool | None = None, _cli_avoid_json: bool | None = None, _cli_enforce_required: bool | None = None, _cli_use_class_docs_for_groups: bool | None = None, _cli_exit_on_error: bool | None = None, _cli_prefix: str | None = None, _cli_flag_prefix_char: str | None = None, _cli_implicit_flags: bool | Literal['dual', 'toggle'] | None = None, _cli_ignore_unknown_args: bool | None = None, _cli_kebab_case: bool | Literal['all', 'no_enums'] | None = None, _cli_shortcuts: Mapping[str, str | list[str]] | None = None, _secrets_dir: PathType | None = None, _build_sources: tuple[tuple[PydanticBaseSettingsSource, ...], dict[str, Any]] | None = None, *, loglevel: int = 10, loglevel_3rdparty: int = 30, our_packages: list[str] = ['__main__', 'qfa'], basicConfig: dict[str, Any] = {'format': '%(asctime)s:%(levelname)s:%(name)s:%(message)s'})[source]#

Bases: BaseSettings

Define settings for the logger.

Create a new model by parsing and validating input data from keyword arguments.

Raises [ValidationError][pydantic_core.ValidationError] if the input data cannot be validated to form a valid model.

self is explicitly positional-only to allow self as a field name.

loglevel: int#

loglevel_3rdparty: int#

our_packages: list[str]#

basicConfig: dict[str, Any]#

classmethod string_to_loglevel(v: str) → int[source]#: Convert a string to a loglevel.

model_config = {'arbitrary_types_allowed': True, 'case_sensitive': False, 'cli_avoid_json': False, 'cli_enforce_required': False, 'cli_exit_on_error': True, 'cli_flag_prefix_char': '-', 'cli_hide_none_type': False, 'cli_ignore_unknown_args': False, 'cli_implicit_flags': False, 'cli_kebab_case': False, 'cli_parse_args': None, 'cli_parse_none_str': None, 'cli_prefix': '', 'cli_prog_name': None, 'cli_shortcuts': None, 'cli_use_class_docs_for_groups': False, 'enable_decoding': True, 'env_file': None, 'env_file_encoding': None, 'env_ignore_empty': False, 'env_nested_delimiter': None, 'env_nested_max_split': None, 'env_parse_enums': None, 'env_parse_none_str': None, 'env_prefix': '', 'env_prefix_target': 'variable', 'extra': 'forbid', 'json_file': None, 'json_file_encoding': None, 'nested_model_default_partial_update': False, 'protected_namespaces': ('model_validate', 'model_dump', 'settings_customise_sources'), 'secrets_dir': None, 'toml_file': None, 'validate_default': True, 'yaml_config_section': None, 'yaml_file': None, 'yaml_file_encoding': None}#: Configuration for the model, should be a dictionary conforming to [ConfigDict][pydantic.config.ConfigDict].

class qfa.settings.LLMSettings(_case_sensitive: bool | None = None, _nested_model_default_partial_update: bool | None = None, _env_prefix: str | None = None, _env_prefix_target: EnvPrefixTarget | None = None, _env_file: DotenvType | None = PosixPath('.'), _env_file_encoding: str | None = None, _env_ignore_empty: bool | None = None, _env_nested_delimiter: str | None = None, _env_nested_max_split: int | None = None, _env_parse_none_str: str | None = None, _env_parse_enums: bool | None = None, _cli_prog_name: str | None = None, _cli_parse_args: bool | list[str] | tuple[str, ...] | None = None, _cli_settings_source: CliSettingsSource[Any] | None = None, _cli_parse_none_str: str | None = None, _cli_hide_none_type: bool | None = None, _cli_avoid_json: bool | None = None, _cli_enforce_required: bool | None = None, _cli_use_class_docs_for_groups: bool | None = None, _cli_exit_on_error: bool | None = None, _cli_prefix: str | None = None, _cli_flag_prefix_char: str | None = None, _cli_implicit_flags: bool | Literal['dual', 'toggle'] | None = None, _cli_ignore_unknown_args: bool | None = None, _cli_kebab_case: bool | Literal['all', 'no_enums'] | None = None, _cli_shortcuts: Mapping[str, str | list[str]] | None = None, _secrets_dir: PathType | None = None, _build_sources: tuple[tuple[PydanticBaseSettingsSource, ...], dict[str, Any]] | None = None, *, model: str = 'azure/gpt-5.4', api_key: SecretStr, api_base: str = '', api_version: str = '', timeout_seconds: float = 230.0, max_total_tokens: int = 100000, chars_per_token: int = 4)[source]#

Bases: BaseSettings

Configuration for the LLM provider connection.

The provider is inferred from the model string prefix by LiteLLM (e.g. "azure/gpt-4" for Azure OpenAI, "azure_ai/mistral-large" for Azure AI serverless endpoints).

Create a new model by parsing and validating input data from keyword arguments.

Raises [ValidationError][pydantic_core.ValidationError] if the input data cannot be validated to form a valid model.

self is explicitly positional-only to allow self as a field name.

model_config = {'arbitrary_types_allowed': True, 'case_sensitive': False, 'cli_avoid_json': False, 'cli_enforce_required': False, 'cli_exit_on_error': True, 'cli_flag_prefix_char': '-', 'cli_hide_none_type': False, 'cli_ignore_unknown_args': False, 'cli_implicit_flags': False, 'cli_kebab_case': False, 'cli_parse_args': None, 'cli_parse_none_str': None, 'cli_prefix': '', 'cli_prog_name': None, 'cli_shortcuts': None, 'cli_use_class_docs_for_groups': False, 'enable_decoding': True, 'env_file': None, 'env_file_encoding': None, 'env_ignore_empty': False, 'env_nested_delimiter': None, 'env_nested_max_split': None, 'env_parse_enums': None, 'env_parse_none_str': None, 'env_prefix': 'LLM_', 'env_prefix_target': 'variable', 'extra': 'forbid', 'json_file': None, 'json_file_encoding': None, 'nested_model_default_partial_update': False, 'protected_namespaces': ('model_validate', 'model_dump', 'settings_customise_sources'), 'secrets_dir': None, 'toml_file': None, 'validate_default': True, 'yaml_config_section': None, 'yaml_file': None, 'yaml_file_encoding': None}#: Configuration for the model, should be a dictionary conforming to [ConfigDict][pydantic.config.ConfigDict].

model: str#

api_key: SecretStr#

api_base: str#

api_version: str#

timeout_seconds: float#

max_total_tokens: int#

chars_per_token: int#

Bases: BaseSettings

Optional per-field overrides for the LLM connection used by judge calls.

Judge calls (the LLM-as-judge quality scores attached to analyse and summarise results) can run on a different model than generation, so the generator does not grade its own homework. This block configures that second connection.

Every field is optional and defaults to ``None``, which means “inherit the corresponding value from the primary LLMSettings “. Deliberately not a subclass of LLMSettings: that would inherit api_key’s required-ness and force a second credential — and therefore a new Key Vault secret in every environment — before a judge model could be tried at all. Here the credential is inherited by default, so enabling a judge on the same Azure resource needs one non-secret variable:

JUDGE_LLM_MODEL=azure_ai/mistral-medium-3-5
JUDGE_LLM_API_BASE=https://<resource>.services.ai.azure.com/models

model is the switch: while it is unset (or empty) no judge client is built at all and judge calls keep using the primary client, so the default configuration behaves exactly as it did before this block existed. The remaining overrides only take effect once model is set.

Note that None (unset) and "" (explicitly set to empty) differ for the string fields: the former inherits the primary’s value, the latter overrides it to empty. timeout_seconds, max_total_tokens and chars_per_token are always taken from the primary settings and have no judge-side override.

Resolution against the primary settings happens once at composition, in qfa.api.composition.resolve_judge_llm_settings().

Create a new model by parsing and validating input data from keyword arguments.

Raises [ValidationError][pydantic_core.ValidationError] if the input data cannot be validated to form a valid model.

self is explicitly positional-only to allow self as a field name.

model_config = {'arbitrary_types_allowed': True, 'case_sensitive': False, 'cli_avoid_json': False, 'cli_enforce_required': False, 'cli_exit_on_error': True, 'cli_flag_prefix_char': '-', 'cli_hide_none_type': False, 'cli_ignore_unknown_args': False, 'cli_implicit_flags': False, 'cli_kebab_case': False, 'cli_parse_args': None, 'cli_parse_none_str': None, 'cli_prefix': '', 'cli_prog_name': None, 'cli_shortcuts': None, 'cli_use_class_docs_for_groups': False, 'enable_decoding': True, 'env_file': None, 'env_file_encoding': None, 'env_ignore_empty': False, 'env_nested_delimiter': None, 'env_nested_max_split': None, 'env_parse_enums': None, 'env_parse_none_str': None, 'env_prefix': 'JUDGE_LLM_', 'env_prefix_target': 'variable', 'extra': 'forbid', 'json_file': None, 'json_file_encoding': None, 'nested_model_default_partial_update': False, 'protected_namespaces': ('model_validate', 'model_dump', 'settings_customise_sources'), 'secrets_dir': None, 'toml_file': None, 'validate_default': True, 'yaml_config_section': None, 'yaml_file': None, 'yaml_file_encoding': None}#: Configuration for the model, should be a dictionary conforming to [ConfigDict][pydantic.config.ConfigDict].

model: str | None#

api_key: SecretStr | None#

api_base: str | None#

api_version: str | None#

class qfa.settings.EmbeddingSettings(_case_sensitive: bool | None = None, _nested_model_default_partial_update: bool | None = None, _env_prefix: str | None = None, _env_prefix_target: EnvPrefixTarget | None = None, _env_file: DotenvType | None = PosixPath('.'), _env_file_encoding: str | None = None, _env_ignore_empty: bool | None = None, _env_nested_delimiter: str | None = None, _env_nested_max_split: int | None = None, _env_parse_none_str: str | None = None, _env_parse_enums: bool | None = None, _cli_prog_name: str | None = None, _cli_parse_args: bool | list[str] | tuple[str, ...] | None = None, _cli_settings_source: CliSettingsSource[Any] | None = None, _cli_parse_none_str: str | None = None, _cli_hide_none_type: bool | None = None, _cli_avoid_json: bool | None = None, _cli_enforce_required: bool | None = None, _cli_use_class_docs_for_groups: bool | None = None, _cli_exit_on_error: bool | None = None, _cli_prefix: str | None = None, _cli_flag_prefix_char: str | None = None, _cli_implicit_flags: bool | Literal['dual', 'toggle'] | None = None, _cli_ignore_unknown_args: bool | None = None, _cli_kebab_case: bool | Literal['all', 'no_enums'] | None = None, _cli_shortcuts: Mapping[str, str | list[str]] | None = None, _secrets_dir: PathType | None = None, _build_sources: tuple[tuple[PydanticBaseSettingsSource, ...], dict[str, Any]] | None = None, *, model_path: str = '', tokenizer_path: str = '', revision_hash: str = '', model_kind: Literal['bge-m3', 'e5'] = 'e5', dense_dim: Annotated[int, Ge(ge=1)] = 768, max_tokens: Annotated[int | None, Ge(ge=1)] = None, intra_op_num_threads: int | None = None, batch_size: Annotated[int, Ge(ge=1)] = 100)[source]#

Bases: BaseSettings

Configuration for the self-hosted embedding model.

The artifact is mirrored locally and pinned by hash; production never fetches it from HuggingFace at runtime.

Create a new model by parsing and validating input data from keyword arguments.

Raises [ValidationError][pydantic_core.ValidationError] if the input data cannot be validated to form a valid model.

self is explicitly positional-only to allow self as a field name.

model_config = {'arbitrary_types_allowed': True, 'case_sensitive': False, 'cli_avoid_json': False, 'cli_enforce_required': False, 'cli_exit_on_error': True, 'cli_flag_prefix_char': '-', 'cli_hide_none_type': False, 'cli_ignore_unknown_args': False, 'cli_implicit_flags': False, 'cli_kebab_case': False, 'cli_parse_args': None, 'cli_parse_none_str': None, 'cli_prefix': '', 'cli_prog_name': None, 'cli_shortcuts': None, 'cli_use_class_docs_for_groups': False, 'enable_decoding': True, 'env_file': None, 'env_file_encoding': None, 'env_ignore_empty': False, 'env_nested_delimiter': None, 'env_nested_max_split': None, 'env_parse_enums': None, 'env_parse_none_str': None, 'env_prefix': 'EMBEDDING_', 'env_prefix_target': 'variable', 'extra': 'forbid', 'json_file': None, 'json_file_encoding': None, 'nested_model_default_partial_update': False, 'protected_namespaces': ('model_validate', 'model_dump', 'settings_customise_sources'), 'secrets_dir': None, 'toml_file': None, 'validate_default': True, 'yaml_config_section': None, 'yaml_file': None, 'yaml_file_encoding': None}#: Configuration for the model, should be a dictionary conforming to [ConfigDict][pydantic.config.ConfigDict].

model_path: str#

tokenizer_path: str#

revision_hash: str#

model_kind: Literal['bge-m3', 'e5']#

dense_dim: int#

max_tokens: int | None#

intra_op_num_threads: int | None#

batch_size: int#

class qfa.settings.OrchestratorSettings(_case_sensitive: bool | None = None, _nested_model_default_partial_update: bool | None = None, _env_prefix: str | None = None, _env_prefix_target: EnvPrefixTarget | None = None, _env_file: DotenvType | None = PosixPath('.'), _env_file_encoding: str | None = None, _env_ignore_empty: bool | None = None, _env_nested_delimiter: str | None = None, _env_nested_max_split: int | None = None, _env_parse_none_str: str | None = None, _env_parse_enums: bool | None = None, _cli_prog_name: str | None = None, _cli_parse_args: bool | list[str] | tuple[str, ...] | None = None, _cli_settings_source: CliSettingsSource[Any] | None = None, _cli_parse_none_str: str | None = None, _cli_hide_none_type: bool | None = None, _cli_avoid_json: bool | None = None, _cli_enforce_required: bool | None = None, _cli_use_class_docs_for_groups: bool | None = None, _cli_exit_on_error: bool | None = None, _cli_prefix: str | None = None, _cli_flag_prefix_char: str | None = None, _cli_implicit_flags: bool | Literal['dual', 'toggle'] | None = None, _cli_ignore_unknown_args: bool | None = None, _cli_kebab_case: bool | Literal['all', 'no_enums'] | None = None, _cli_shortcuts: Mapping[str, str | list[str]] | None = None, _secrets_dir: PathType | None = None, _build_sources: tuple[tuple[PydanticBaseSettingsSource, ...], dict[str, Any]] | None = None, *, metadata_fields_to_include: list[str] = <factory>, retry_base_seconds: float = 1.0, retry_multiplier: float = 2.0, retry_jitter_factor: float = 0.5, retry_cap_seconds: float = 10.0, chars_per_token: int = 4)[source]#

Bases: BaseSettings

Cross-cutting configuration shared by every orchestrator use case.

Per-endpoint tuning lives in its own settings class (e.g. AnalyzeSettings) so the eventual split into one-use-case-per-module (per ADR-011) doesn’t require renaming env-vars in production. Only knobs that genuinely apply to every use case — retry policy, token-budget estimation, metadata allow-list — stay here.

Create a new model by parsing and validating input data from keyword arguments.

Raises [ValidationError][pydantic_core.ValidationError] if the input data cannot be validated to form a valid model.

self is explicitly positional-only to allow self as a field name.

model_config = {'arbitrary_types_allowed': True, 'case_sensitive': False, 'cli_avoid_json': False, 'cli_enforce_required': False, 'cli_exit_on_error': True, 'cli_flag_prefix_char': '-', 'cli_hide_none_type': False, 'cli_ignore_unknown_args': False, 'cli_implicit_flags': False, 'cli_kebab_case': False, 'cli_parse_args': None, 'cli_parse_none_str': None, 'cli_prefix': '', 'cli_prog_name': None, 'cli_shortcuts': None, 'cli_use_class_docs_for_groups': False, 'enable_decoding': True, 'env_file': None, 'env_file_encoding': None, 'env_ignore_empty': False, 'env_nested_delimiter': None, 'env_nested_max_split': None, 'env_parse_enums': None, 'env_parse_none_str': None, 'env_prefix': 'ORCHESTRATOR_', 'env_prefix_target': 'variable', 'extra': 'forbid', 'json_file': None, 'json_file_encoding': None, 'nested_model_default_partial_update': False, 'protected_namespaces': ('model_validate', 'model_dump', 'settings_customise_sources'), 'secrets_dir': None, 'toml_file': None, 'validate_default': True, 'yaml_config_section': None, 'yaml_file': None, 'yaml_file_encoding': None}#: Configuration for the model, should be a dictionary conforming to [ConfigDict][pydantic.config.ConfigDict].

metadata_fields_to_include: list[str]#

retry_base_seconds: float#

retry_multiplier: float#

retry_jitter_factor: float#

retry_cap_seconds: float#

chars_per_token: int#

class qfa.settings.AnalyzeSettings(_case_sensitive: bool | None = None, _nested_model_default_partial_update: bool | None = None, _env_prefix: str | None = None, _env_prefix_target: EnvPrefixTarget | None = None, _env_file: DotenvType | None = PosixPath('.'), _env_file_encoding: str | None = None, _env_ignore_empty: bool | None = None, _env_nested_delimiter: str | None = None, _env_nested_max_split: int | None = None, _env_parse_none_str: str | None = None, _env_parse_enums: bool | None = None, _cli_prog_name: str | None = None, _cli_parse_args: bool | list[str] | tuple[str, ...] | None = None, _cli_settings_source: CliSettingsSource[Any] | None = None, _cli_parse_none_str: str | None = None, _cli_hide_none_type: bool | None = None, _cli_avoid_json: bool | None = None, _cli_enforce_required: bool | None = None, _cli_use_class_docs_for_groups: bool | None = None, _cli_exit_on_error: bool | None = None, _cli_prefix: str | None = None, _cli_flag_prefix_char: str | None = None, _cli_implicit_flags: bool | Literal['dual', 'toggle'] | None = None, _cli_ignore_unknown_args: bool | None = None, _cli_kebab_case: bool | Literal['all', 'no_enums'] | None = None, _cli_shortcuts: Mapping[str, str | list[str]] | None = None, _secrets_dir: PathType | None = None, _build_sources: tuple[tuple[PydanticBaseSettingsSource, ...], dict[str, Any]] | None = None, *, min_cluster_size: ~typing.Annotated[int, ~annotated_types.Ge(ge=2)] = 5, clustering_metric: str = 'euclidean', max_concurrent_chunks: ~typing.Annotated[int, ~annotated_types.Ge(ge=1)] = 64, target_chunk_tokens: ~typing.Annotated[int, ~annotated_types.Ge(ge=1)] = 2000, coding_trend_code_fields: list[str] = <factory>, default_coding_trend_period: ~typing.Literal['day', 'week', 'month'] = 'week')[source]#

Bases: BaseSettings

Configuration specific to the POST /v1/analyze endpoint.

Covers both mode=single_pass and mode=hierarchical: the coding-trend table is built for both, and the clustering knobs are only consulted on the hierarchical path. Naming the group after the endpoint (not the mode) lets a future single_pass-only knob land here without another rename.

Create a new model by parsing and validating input data from keyword arguments.

Raises [ValidationError][pydantic_core.ValidationError] if the input data cannot be validated to form a valid model.

self is explicitly positional-only to allow self as a field name.

model_config = {'arbitrary_types_allowed': True, 'case_sensitive': False, 'cli_avoid_json': False, 'cli_enforce_required': False, 'cli_exit_on_error': True, 'cli_flag_prefix_char': '-', 'cli_hide_none_type': False, 'cli_ignore_unknown_args': False, 'cli_implicit_flags': False, 'cli_kebab_case': False, 'cli_parse_args': None, 'cli_parse_none_str': None, 'cli_prefix': '', 'cli_prog_name': None, 'cli_shortcuts': None, 'cli_use_class_docs_for_groups': False, 'enable_decoding': True, 'env_file': None, 'env_file_encoding': None, 'env_ignore_empty': False, 'env_nested_delimiter': None, 'env_nested_max_split': None, 'env_parse_enums': None, 'env_parse_none_str': None, 'env_prefix': 'ANALYZE_', 'env_prefix_target': 'variable', 'extra': 'forbid', 'json_file': None, 'json_file_encoding': None, 'nested_model_default_partial_update': False, 'protected_namespaces': ('model_validate', 'model_dump', 'settings_customise_sources'), 'secrets_dir': None, 'toml_file': None, 'validate_default': True, 'yaml_config_section': None, 'yaml_file': None, 'yaml_file_encoding': None}#: Configuration for the model, should be a dictionary conforming to [ConfigDict][pydantic.config.ConfigDict].

min_cluster_size: int#

clustering_metric: str#

max_concurrent_chunks: int#

target_chunk_tokens: int#

coding_trend_code_fields: list[str]#

default_coding_trend_period: Literal['day', 'week', 'month']#

Bases: BaseSettings

Configuration for API-key based authentication.

Create a new model by parsing and validating input data from keyword arguments.

Raises [ValidationError][pydantic_core.ValidationError] if the input data cannot be validated to form a valid model.

self is explicitly positional-only to allow self as a field name.

model_config = {'arbitrary_types_allowed': True, 'case_sensitive': False, 'cli_avoid_json': False, 'cli_enforce_required': False, 'cli_exit_on_error': True, 'cli_flag_prefix_char': '-', 'cli_hide_none_type': False, 'cli_ignore_unknown_args': False, 'cli_implicit_flags': False, 'cli_kebab_case': False, 'cli_parse_args': None, 'cli_parse_none_str': None, 'cli_prefix': '', 'cli_prog_name': None, 'cli_shortcuts': None, 'cli_use_class_docs_for_groups': False, 'enable_decoding': True, 'env_file': None, 'env_file_encoding': None, 'env_ignore_empty': False, 'env_nested_delimiter': None, 'env_nested_max_split': None, 'env_parse_enums': None, 'env_parse_none_str': None, 'env_prefix': 'AUTH_', 'env_prefix_target': 'variable', 'extra': 'forbid', 'json_file': None, 'json_file_encoding': None, 'nested_model_default_partial_update': False, 'protected_namespaces': ('model_validate', 'model_dump', 'settings_customise_sources'), 'secrets_dir': None, 'toml_file': None, 'validate_default': True, 'yaml_config_section': None, 'yaml_file': None, 'yaml_file_encoding': None}#: Configuration for the model, should be a dictionary conforming to [ConfigDict][pydantic.config.ConfigDict].

api_keys: list[TenantApiKey]#

class qfa.settings.DatabaseSettings(_case_sensitive: bool | None = None, _nested_model_default_partial_update: bool | None = None, _env_prefix: str | None = None, _env_prefix_target: EnvPrefixTarget | None = None, _env_file: DotenvType | None = PosixPath('.'), _env_file_encoding: str | None = None, _env_ignore_empty: bool | None = None, _env_nested_delimiter: str | None = None, _env_nested_max_split: int | None = None, _env_parse_none_str: str | None = None, _env_parse_enums: bool | None = None, _cli_prog_name: str | None = None, _cli_parse_args: bool | list[str] | tuple[str, ...] | None = None, _cli_settings_source: CliSettingsSource[Any] | None = None, _cli_parse_none_str: str | None = None, _cli_hide_none_type: bool | None = None, _cli_avoid_json: bool | None = None, _cli_enforce_required: bool | None = None, _cli_use_class_docs_for_groups: bool | None = None, _cli_exit_on_error: bool | None = None, _cli_prefix: str | None = None, _cli_flag_prefix_char: str | None = None, _cli_implicit_flags: bool | Literal['dual', 'toggle'] | None = None, _cli_ignore_unknown_args: bool | None = None, _cli_kebab_case: bool | Literal['all', 'no_enums'] | None = None, _cli_shortcuts: Mapping[str, str | list[str]] | None = None, _secrets_dir: PathType | None = None, _build_sources: tuple[tuple[PydanticBaseSettingsSource, ...], dict[str, Any]] | None = None, *, url: str = '', host: str = '', port: int = 5432, name: str = '', user: str = '', password: SecretStr | None = None, auth_mode: Literal['password', 'entra'] = 'password', aad_scope: str = 'https://ossrdbms-aad.database.windows.net/.default')[source]#

Bases: BaseSettings

Configuration for the PostgreSQL database connection.

Variables:: url (str) – Database connection URL (asyncpg dialect).

Create a new model by parsing and validating input data from keyword arguments.

Raises [ValidationError][pydantic_core.ValidationError] if the input data cannot be validated to form a valid model.

self is explicitly positional-only to allow self as a field name.

model_config = {'arbitrary_types_allowed': True, 'case_sensitive': False, 'cli_avoid_json': False, 'cli_enforce_required': False, 'cli_exit_on_error': True, 'cli_flag_prefix_char': '-', 'cli_hide_none_type': False, 'cli_ignore_unknown_args': False, 'cli_implicit_flags': False, 'cli_kebab_case': False, 'cli_parse_args': None, 'cli_parse_none_str': None, 'cli_prefix': '', 'cli_prog_name': None, 'cli_shortcuts': None, 'cli_use_class_docs_for_groups': False, 'enable_decoding': True, 'env_file': None, 'env_file_encoding': None, 'env_ignore_empty': False, 'env_nested_delimiter': None, 'env_nested_max_split': None, 'env_parse_enums': None, 'env_parse_none_str': None, 'env_prefix': 'DB_', 'env_prefix_target': 'variable', 'extra': 'forbid', 'json_file': None, 'json_file_encoding': None, 'nested_model_default_partial_update': False, 'protected_namespaces': ('model_validate', 'model_dump', 'settings_customise_sources'), 'secrets_dir': None, 'toml_file': None, 'validate_default': True, 'yaml_config_section': None, 'yaml_file': None, 'yaml_file_encoding': None}#: Configuration for the model, should be a dictionary conforming to [ConfigDict][pydantic.config.ConfigDict].

url: str#

host: str#

port: int#

name: str#

user: str#

password: SecretStr | None#

auth_mode: Literal['password', 'entra']#

aad_scope: str#

Bases: BaseSettings

Configuration for network settings.

Create a new model by parsing and validating input data from keyword arguments.

Raises [ValidationError][pydantic_core.ValidationError] if the input data cannot be validated to form a valid model.

self is explicitly positional-only to allow self as a field name.

model_config = {'arbitrary_types_allowed': True, 'case_sensitive': False, 'cli_avoid_json': False, 'cli_enforce_required': False, 'cli_exit_on_error': True, 'cli_flag_prefix_char': '-', 'cli_hide_none_type': False, 'cli_ignore_unknown_args': False, 'cli_implicit_flags': False, 'cli_kebab_case': False, 'cli_parse_args': None, 'cli_parse_none_str': None, 'cli_prefix': '', 'cli_prog_name': None, 'cli_shortcuts': None, 'cli_use_class_docs_for_groups': False, 'enable_decoding': True, 'env_file': None, 'env_file_encoding': None, 'env_ignore_empty': False, 'env_nested_delimiter': None, 'env_nested_max_split': None, 'env_parse_enums': None, 'env_parse_none_str': None, 'env_prefix': 'NETWORK_', 'env_prefix_target': 'variable', 'extra': 'forbid', 'json_file': None, 'json_file_encoding': None, 'nested_model_default_partial_update': False, 'protected_namespaces': ('model_validate', 'model_dump', 'settings_customise_sources'), 'secrets_dir': None, 'toml_file': None, 'validate_default': True, 'yaml_config_section': None, 'yaml_file': None, 'yaml_file_encoding': None}#: Configuration for the model, should be a dictionary conforming to [ConfigDict][pydantic.config.ConfigDict].

host: str#

port: int#

Bases: BaseSettings

Azure Monitor / Application Insights telemetry configuration.

Kept as its own group with no required fields so it can be constructed standalone at import time (qfa.main reads it before building the full application graph) without tripping the required env-vars of the other settings groups.

Create a new model by parsing and validating input data from keyword arguments.

Raises [ValidationError][pydantic_core.ValidationError] if the input data cannot be validated to form a valid model.

self is explicitly positional-only to allow self as a field name.

model_config = {'arbitrary_types_allowed': True, 'case_sensitive': False, 'cli_avoid_json': False, 'cli_enforce_required': False, 'cli_exit_on_error': True, 'cli_flag_prefix_char': '-', 'cli_hide_none_type': False, 'cli_ignore_unknown_args': False, 'cli_implicit_flags': False, 'cli_kebab_case': False, 'cli_parse_args': None, 'cli_parse_none_str': None, 'cli_prefix': '', 'cli_prog_name': None, 'cli_shortcuts': None, 'cli_use_class_docs_for_groups': False, 'enable_decoding': True, 'env_file': None, 'env_file_encoding': None, 'env_ignore_empty': False, 'env_nested_delimiter': None, 'env_nested_max_split': None, 'env_parse_enums': None, 'env_parse_none_str': None, 'env_prefix': '', 'env_prefix_target': 'variable', 'extra': 'forbid', 'json_file': None, 'json_file_encoding': None, 'nested_model_default_partial_update': False, 'protected_namespaces': ('model_validate', 'model_dump', 'settings_customise_sources'), 'secrets_dir': None, 'toml_file': None, 'validate_default': True, 'yaml_config_section': None, 'yaml_file': None, 'yaml_file_encoding': None}#: Configuration for the model, should be a dictionary conforming to [ConfigDict][pydantic.config.ConfigDict].

applicationinsights_connection_string: SecretStr | None#

Azure Application Insights connection string.

Read from APPLICATIONINSIGHTS_CONNECTION_STRING (set on the App Service by infra/app_service.tf). When present, qfa.main initialises the Azure Monitor OpenTelemetry SDK with it; unset in local dev, which leaves telemetry export disabled. Held as SecretStr — it embeds an ingestion key — so it is masked in logs and model_dump output.

class qfa.settings.AppSettings(_case_sensitive: bool | None = None, _nested_model_default_partial_update: bool | None = None, _env_prefix: str | None = None, _env_prefix_target: EnvPrefixTarget | None = None, _env_file: DotenvType | None = PosixPath('.'), _env_file_encoding: str | None = None, _env_ignore_empty: bool | None = None, _env_nested_delimiter: str | None = None, _env_nested_max_split: int | None = None, _env_parse_none_str: str | None = None, _env_parse_enums: bool | None = None, _cli_prog_name: str | None = None, _cli_parse_args: bool | list[str] | tuple[str, ...] | None = None, _cli_settings_source: CliSettingsSource[Any] | None = None, _cli_parse_none_str: str | None = None, _cli_hide_none_type: bool | None = None, _cli_avoid_json: bool | None = None, _cli_enforce_required: bool | None = None, _cli_use_class_docs_for_groups: bool | None = None, _cli_exit_on_error: bool | None = None, _cli_prefix: str | None = None, _cli_flag_prefix_char: str | None = None, _cli_implicit_flags: bool | Literal['dual', 'toggle'] | None = None, _cli_ignore_unknown_args: bool | None = None, _cli_kebab_case: bool | Literal['all', 'no_enums'] | None = None, _cli_shortcuts: Mapping[str, str | list[str]] | None = None, _secrets_dir: PathType | None = None, _build_sources: tuple[tuple[PydanticBaseSettingsSource, ...], dict[str, Any]] | None = None, *, llm: ~qfa.settings.LLMSettings = <factory>, judge_llm: ~qfa.settings.JudgeLLMSettings = <factory>, embedding: ~qfa.settings.EmbeddingSettings = <factory>, orchestrator: ~qfa.settings.OrchestratorSettings = <factory>, analyze: ~qfa.settings.AnalyzeSettings = <factory>, auth: ~qfa.settings.AuthSettings = <factory>, log: ~qfa.settings.LogSettings = <factory>, db: ~qfa.settings.DatabaseSettings = <factory>, network: ~qfa.settings.NetworkSettings = <factory>, telemetry: ~qfa.settings.TelemetrySettings = <factory>, debug: bool = False)[source]#

Bases: BaseSettings

Root configuration composing all sub-settings groups.

Create a new model by parsing and validating input data from keyword arguments.

Raises [ValidationError][pydantic_core.ValidationError] if the input data cannot be validated to form a valid model.

self is explicitly positional-only to allow self as a field name.

llm: LLMSettings#

judge_llm: JudgeLLMSettings#

embedding: EmbeddingSettings#

orchestrator: OrchestratorSettings#

analyze: AnalyzeSettings#

auth: AuthSettings#

log: LogSettings#

db: DatabaseSettings#

network: NetworkSettings#

model_config = {'arbitrary_types_allowed': True, 'case_sensitive': False, 'cli_avoid_json': False, 'cli_enforce_required': False, 'cli_exit_on_error': True, 'cli_flag_prefix_char': '-', 'cli_hide_none_type': False, 'cli_ignore_unknown_args': False, 'cli_implicit_flags': False, 'cli_kebab_case': False, 'cli_parse_args': None, 'cli_parse_none_str': None, 'cli_prefix': '', 'cli_prog_name': None, 'cli_shortcuts': None, 'cli_use_class_docs_for_groups': False, 'enable_decoding': True, 'env_file': None, 'env_file_encoding': None, 'env_ignore_empty': False, 'env_nested_delimiter': None, 'env_nested_max_split': None, 'env_parse_enums': None, 'env_parse_none_str': None, 'env_prefix': '', 'env_prefix_target': 'variable', 'extra': 'forbid', 'json_file': None, 'json_file_encoding': None, 'nested_model_default_partial_update': False, 'protected_namespaces': ('model_validate', 'model_dump', 'settings_customise_sources'), 'secrets_dir': None, 'toml_file': None, 'validate_default': True, 'yaml_config_section': None, 'yaml_file': None, 'yaml_file_encoding': None}#: Configuration for the model, should be a dictionary conforming to [ConfigDict][pydantic.config.ConfigDict].

telemetry: TelemetrySettings#

debug: bool#

Whether to enable debug mode.

This will, e.g., enable code reloading for the uvicorn server.