Testing & Benchmarks

SWE_BENCH_INSTANCE_ID

• Type: String (instance identifier)
• Default: Unspecified
• Description: SWE-bench instance identifier. Used in SWE-bench benchmarking runs.
• Example: export SWE_BENCH_INSTANCE_ID=instance-123

SWE_BENCH_RUN_ID

• Type: String (run identifier)
• Default: Unspecified
• Description: SWE-bench run identifier for benchmarking. Tracks a benchmarking run.
• Example: export SWE_BENCH_RUN_ID=run-abc123

SWE_BENCH_TASK_ID

• Type: String (task identifier)
• Default: Unspecified
• Description: SWE-bench task identifier. Identifies the specific task being benchmarked.
• Example: export SWE_BENCH_TASK_ID=task-xyz789

CLAUDE_CODE_TEST_FIXTURES_ROOT

• Type: String (directory path)
• Default: Unspecified
• Description: Root directory for test fixtures. Used in testing to locate test data.
• Example: export CLAUDE_CODE_TEST_FIXTURES_ROOT=/test/fixtures

TEST_ENABLE_SESSION_PERSISTENCE

• Type: Boolean (1, true, 0, false)
• Default: false
• Description: Enable session persistence in tests. Allows sessions to persist across test runs.
• Example: export TEST_ENABLE_SESSION_PERSISTENCE=1

TEST_GRACEFUL_FS_GLOBAL_PATCH

• Type: Boolean (1, true, 0, false)
• Default: false
• Description: Enable graceful-fs global patching in tests. Patches filesystem operations for test stability.
• Example: export TEST_GRACEFUL_FS_GLOBAL_PATCH=1

VCR_RECORD

• Type: String (record mode)
• Default: Unspecified
• Description: VCR recording mode for test replay. Controls whether HTTP interactions are recorded or replayed.
• Example: export VCR_RECORD=once

AWS_LAMBDA_BENCHMARK_MODE

• Type: Boolean (1, true, 0, false)
• Default: false
• Description: Set to ‘1’ to enable Lambda benchmark testing mode.
• Example: export AWS_LAMBDA_BENCHMARK_MODE=1

---

← Back to env/README.md