# TeaQL safe SQL execution evidence (Python) Collect provider-native parameterized SQL only on a trusted server-created `UserContext`. Convert trusted log entries to this bounded shape before exposing diagnostics to an application or Coding Agent. ```python from dataclasses import dataclass from typing import Optional, Sequence from teaql.runtime import SqlLogEntry, SqlLogOperation from teaql.core import TraceNode @dataclass(frozen=True) class SafeSQLExecutionEvidence: operation: SqlLogOperation parameterized_sql: str parameter_count: int elapsed_micros: int result_count: Optional[int] affected_rows: Optional[int] result_summary: str trace_chain: tuple[TraceNode, ...] trace_path: tuple[TraceNode, ...] def safe_sql_evidence(entries: Sequence[SqlLogEntry]) -> list[SafeSQLExecutionEvidence]: return [ SafeSQLExecutionEvidence( operation=entry.operation, parameterized_sql=entry.sql, parameter_count=len(entry.params), elapsed_micros=int(entry.elapsed.total_seconds() * 1_000_000), result_count=entry.result_count, affected_rows=entry.affected_rows, result_summary=entry.result_summary, trace_chain=tuple(entry.trace_chain), trace_path=tuple(entry.trace_path), ) for entry in entries ] ``` Use `enable_all_sql_log()`, `enable_select_sql_log()`, `enable_mutation_sql_log()`, or `disable_sql_log()` on the trusted context. Every mode change clears previous entries. Raw `params`, interpolated `debug_sql`, credentials, and connection strings stay inside the trusted boundary and are intentionally absent from the safe projection. Query and mutation logs and `TextDiagnosticSqlLogSink` are enabled on a new context. Replace it, when needed, with `context.with_diagnostic_sql_log_sink(TextDiagnosticSqlLogSink())`. Use `disable_select_sql_log()` and `disable_mutation_sql_log()` independently; use `disable_sql_log()` only to silence both. Each entry retains structured `comment`, `purpose`, `audit_reason`, business `trace_chain`, physical `trace_path`, parameterized SQL, copy-paste `debug_sql`, elapsed time, result count, and affected rows. The business chain preserves the root purpose, native Q-operation parents and each nested query's explicit/default comment. Do not reconstruct it from SQL or the physical route. Log entries are policy-projected by default; keep that projection when exposing evidence. Explicit debug plaintext remains inside the trusted debugging boundary. --- ## TeaQL seven-language assist contract Apply the verified Rust semantic ceiling while using only the exact PYTHON generated and runtime APIs. Discover APIs through the generated application AGENTS.md and progressive model-aware Assist. Do not inspect generated domain-library source. - Do not create plurals by appending `s` or `es`; use the centralized generated plural. - Human and non-human entities use different generated predicate vocabularies. Preserve forms such as “who are active” and “whose email is”; never infer them from English. - Configure filters, projection, paging, and other query options before `purpose(...)`. Comment may appear anywhere in the chain. Purpose enters the executable stage; execution requires both values, but comment does not have to immediately precede purpose. - Every execute/list/stream and every save accepts exactly one context argument: `UserContext`. Name that argument `context`, never `runtime`; data services and global policy are injected when the context is built. Reserve `runtime` for process-level runtime ownership, provider/pool setup, and module assembly. - Tenant, merchant, identity, permissions, request policy, purpose policy, hard limit, and continuous-page cursor policy come only from trusted context, never dynamic JSON or TFP. - If the required operation is absent after current entity/action and required field Assist, stop that path and report MISSING_ASSIST. Do not guess an API or search the generated library as a fallback. - Create each application-owned source file once. After its first compile attempt, repair only the smallest block identified by the exact compiler or test diagnostic. Preserve unrelated code; do not rewrite the complete file as an error-recovery loop. - Before a repair that would replace more than 25% of an existing application file, stop and report LARGE_REWRITE_REQUEST with the file, exact diagnostic, reason, and estimated scope. Initial creation and model-driven regeneration are not repairs. Capability: `debug`. - Capture purpose, comment, trace/correlation id, runtime-projected expanded masked SQL and its masking mode, duration, row count, provider and the native response. - Preserve the immutable row audit event and the customizable App Audit Sink as separate paths. Redact credentials, tokens, connection strings, and customer data. - Document only switches and hooks present in the selected runtime source. ## Standard OpenTelemetry module Use the optional standard module supplied by the selected TeaQL runtime. Match its version to the runtime dependency selected by the model and workspace; keep local source or internal Registry overrides during development. Do not upgrade to an unrelated public release just to enable telemetry. A runtime without this module reports `module_unavailable`; package installation alone does not enable export. On a supported backend, configure these trusted process variables before startup, never through a query, JSON or TFP payload: ```bash export TEAQL_TELEMETRY_ENABLED=true export TEAQL_TELEMETRY_SERVICE_NAME=payment-service export TEAQL_TELEMETRY_OTLP_ENDPOINT=http://localhost:4318 ``` Without enablement, startup reports `disabled` and all three activation variables; no SDK export resources are created. Enabled startup reports safe effective settings, full recording and masking. Invalid settings produce a fixed safe diagnostic, not a claim of successful export. Never print credentials or raw configuration values. The module automatically records supported Q, graph mutation and physical SQL operations as traces, logs and metrics. SQL is expanded with masked parameters; plaintext debugging never authorizes plaintext telemetry export. Full recording does not promise lossless delivery when bounded queues or exporters fail. Do not add per-method spans or replace the customer's global SDK providers. Keep schema reconciliation explicit. Run the generated ensure-schema and seed flow through the same configured runtime to observe startup settings and audited, idempotent seeding. Telemetry setup must not run ensure-schema implicitly. Flush and shut down the owned pipeline at application shutdown, not per query. Exporter failures do not change business results; durable business audit remains an independent path. ### Python assembly Select the `telemetry` extra on the application's existing `teaql` dependency (`teaql[telemetry]`), retaining the version and source or Registry selected by the workspace. UserContext startup loads the module only when enabled. No application SDK initialization or per-method spans are required; set the three variables before starting the process.