You have access to a SQLite database containing CI/CD build, test, and timeline data from Azure DevOps. The database is located at ~/.tiger/tiger.db. Use sqlite3 ~/.tiger/tiger.db to query it.

Database Schema

builds

Completed CI builds from Azure DevOps.

Column	Type	Description
organization	TEXT	AzDO organization (e.g. "dnceng-public")
project	TEXT	AzDO project (e.g. "public")
build_id	INTEGER	Unique build ID
build_number	TEXT	Build number string
definition_name	TEXT	Pipeline/definition name
definition_id	INTEGER	Pipeline definition ID
status	TEXT	Build status (always "completed")
result	TEXT	"succeeded", "failed", "partiallySucceeded", "canceled"
source_branch	TEXT	Branch (e.g. "refs/heads/main", "refs/pull/123/merge")
source_version	TEXT	Git commit SHA
repository_name	TEXT	Repository (e.g. "dotnet/roslyn")
pr_number	INTEGER	PR number if this is a PR build, NULL otherwise
finish_time	TEXT	ISO 8601 finish time
ingested_at	TEXT	When the build was added to the DB

Primary key: (organization, build_id) — build IDs are unique within an organization

test_runs

Test run summaries per build. A build can have multiple test runs (one per job/leg).

Column	Type	Description
organization	TEXT	AzDO organization
project	TEXT	AzDO project
build_id	INTEGER	FK to builds
run_id	INTEGER	Test run ID
run_name	TEXT	Test run/job name
total_tests	INTEGER	Total test count
passed_tests	INTEGER	Passed count
failed_tests	INTEGER	Failed count
skipped_tests	INTEGER	Skipped count

Primary key: (organization, run_id)

test_results

Individual test failure results. Only failed tests are stored.

Column	Type	Description
organization	TEXT	AzDO organization
project	TEXT	AzDO project
run_id	INTEGER	FK to test_runs
result_id	INTEGER	Test result ID
test_case_title	TEXT	Fully qualified test name
outcome	TEXT	Always "Failed" (only failures stored)
error_message	TEXT	Error message (may be NULL)
stack_trace	TEXT	Stack trace (may be NULL)
helix_job_name	TEXT	Helix job ID if test ran on Helix
helix_work_item_name	TEXT	Helix work item name if applicable
is_helix_work_item	INTEGER	1 if this is a synthetic test result created by Helix infrastructure (not a real test)

Primary key: (organization, run_id, result_id)

build_timeline_issues

Errors and warnings from the AzDO build timeline (jobs and tasks).

Column	Type	Description
organization	TEXT	AzDO organization
build_id	INTEGER	FK to builds
record_name	TEXT	Job or task name that produced the issue
record_type	TEXT	"Job" or "Task"
parent_name	TEXT	Parent job name (for Task records)
record_result	TEXT	"failed", "succeeded", etc.
issue_type	TEXT	"error" or "warning"
issue_message	TEXT	The error/warning message
issue_category	TEXT	Issue category (may be NULL)
log_url	TEXT	AzDO timeline log URL for the record (may be NULL)

build_ingestion_tasks

Tracks async ingestion of detailed data per build.

Column	Type	Description
organization	TEXT	AzDO organization
build_id	INTEGER	FK to builds
task_type	TEXT	"tests", "timeline", or "pr_info"
status	TEXT	"pending", "running", "complete", "failed", "abandoned"
is_complete	INTEGER	1 when task is terminal (complete or abandoned), 0 otherwise. A build is fully ingested when all its tasks have is_complete = 1.
attempts	INTEGER	Number of attempts so far
last_error	TEXT	Last error message if failed
last_attempt_time	TEXT	ISO 8601 time of last attempt
next_retry_time	TEXT	ISO 8601 time when a failed task can be retried
completed_time	TEXT	ISO 8601 time when the task completed

Primary key: (organization, build_id, task_type)

pull_requests

Cached PR metadata fetched from GitHub.

Column	Type	Description
repository	TEXT	GitHub repository (e.g. "dotnet/roslyn")
pr_number	INTEGER	Pull request number
title	TEXT	PR title
author	TEXT	PR author login
fetched_at	TEXT	When the PR info was cached

Primary key: (repository, pr_number)

known_issues

Known build errors from GitHub issues labeled "Known Build Error". Refreshed every 15 minutes.

Column	Type	Description
repository	TEXT	GitHub repository
issue_number	INTEGER	GitHub issue number
title	TEXT	Issue title
error_message	TEXT	String or JSON array for contains matching
error_pattern	TEXT	Regex string or JSON array for pattern matching
build_retry	INTEGER	1 if build should be retried
exclude_console_log	INTEGER	1 to skip console log matching
state	TEXT	"open" or "closed"
closed_at	TEXT	When the issue was detected as closed
fetched_at	TEXT	Last refresh time

Primary key: (repository, issue_number)

helix_work_items

Cached Helix work item details for failed tests that ran on Helix.

Column	Type	Description
job_name	TEXT	Helix job ID
work_item_name	TEXT	Helix work item name
state	TEXT	Work item state (e.g. "Finished")
exit_code	INTEGER	Process exit code (NULL if not finished)
console_output_uri	TEXT	URI to console output log
files	TEXT	JSON array of attached files (crash dumps, etc.), excluding the console log. Each entry has `fileName` and `uri`. NULL if no extra files.
is_deadletter	INTEGER	1 if the work item was dead-lettered (Helix infrastructure failure, not a real test failure)

Primary key: (job_name, work_item_name)

agent_tasks

Tracks Copilot Coding Agent tasks submitted from Tiger via gh agent-task create.

Column	Type	Description
session_id	TEXT	Agent session GUID (primary key)
repository	TEXT	GitHub repository (owner/repo format)
test_name	TEXT	Test case title that triggered the task (NULL if not test-related)
file_path	TEXT	Path to the markdown task file on disk
created_at	TEXT	ISO 8601 timestamp when the task was submitted

Primary key: (session_id)

Common Queries

Recent failed builds

SELECT build_id, definition_name, result, repository_name, finish_time
FROM builds
WHERE result = 'failed'
ORDER BY finish_time DESC
LIMIT 20;

Top failing tests (by number of builds affected)

SELECT tr.test_case_title, COUNT(DISTINCT r.build_id) as build_count
FROM test_results tr
JOIN test_runs r ON tr.organization = r.organization
    AND tr.run_id = r.run_id
WHERE tr.outcome = 'Failed'
GROUP BY tr.test_case_title
ORDER BY build_count DESC
LIMIT 20;

Failed tests for a specific build

SELECT tr.test_case_title, tr.error_message, tr.helix_job_name
FROM test_results tr
JOIN test_runs r ON tr.organization = r.organization
    AND tr.run_id = r.run_id
WHERE r.build_id = BUILD_ID AND tr.outcome = 'Failed'
ORDER BY tr.test_case_title;

Timeline errors for a build (failed jobs and their errors)

SELECT parent_name, record_name, issue_type, issue_message
FROM build_timeline_issues
WHERE build_id = BUILD_ID AND issue_type = 'error'
ORDER BY parent_name, record_name;

Builds where a specific test failed

SELECT b.build_id, b.definition_name, b.result, b.finish_time, b.source_branch
FROM test_results tr
JOIN test_runs r ON tr.organization = r.organization
    AND tr.run_id = r.run_id
JOIN builds b ON r.organization = b.organization
    AND r.build_id = b.build_id
WHERE tr.test_case_title = 'FULL_TEST_NAME' AND tr.outcome = 'Failed'
ORDER BY b.finish_time DESC;

Tests that fail on a specific branch

SELECT tr.test_case_title, COUNT(*) as fail_count
FROM test_results tr
JOIN test_runs r ON tr.organization = r.organization
    AND tr.run_id = r.run_id
JOIN builds b ON r.organization = b.organization
    AND r.build_id = b.build_id
WHERE tr.outcome = 'Failed' AND b.source_branch = 'refs/heads/main'
GROUP BY tr.test_case_title
ORDER BY fail_count DESC
LIMIT 20;

Failed builds for a specific repo

SELECT build_id, definition_name, result, finish_time, pr_number
FROM builds
WHERE repository_name = 'dotnet/roslyn' AND result = 'failed'
ORDER BY finish_time DESC
LIMIT 20;

Ingestion status summary

SELECT task_type, status, COUNT(*) as count
FROM build_ingestion_tasks
GROUP BY task_type, status
ORDER BY task_type, status;

Helix work items for a failed test

SELECT tr.test_case_title, tr.helix_job_name, tr.helix_work_item_name, hw.state, hw.exit_code
FROM test_results tr
LEFT JOIN helix_work_items hw ON tr.helix_job_name = hw.job_name
    AND tr.helix_work_item_name = hw.work_item_name
WHERE tr.helix_job_name IS NOT NULL AND tr.outcome = 'Failed'
ORDER BY tr.test_case_title
LIMIT 20;

The Helix console log URL can be constructed as: https://helix.dot.net/api/2019-06-17/jobs/{helix_job_name}/workitems/{helix_work_item_name}/console

PR info for a build

SELECT b.build_id, b.definition_name, b.pr_number, pr.title, pr.author
FROM builds b
LEFT JOIN pull_requests pr ON b.repository_name = pr.repository AND b.pr_number = pr.pr_number
WHERE b.pr_number IS NOT NULL
ORDER BY b.finish_time DESC
LIMIT 20;

Known issues for a repository

SELECT issue_number, title, state, error_message, error_pattern
FROM known_issues
WHERE repository = 'dotnet/roslyn' AND state = 'open'
ORDER BY issue_number;

All builds for a specific PR

SELECT b.build_id, b.definition_name, b.result, b.finish_time
FROM builds b
WHERE b.repository_name = 'dotnet/roslyn' AND b.pr_number = 12345
ORDER BY b.finish_time DESC;

Tips

All build/test tables are keyed by (organization, project) for multi-org support
pull_requests, known_issues, and helix_work_items are keyed by repository or job (not org/project)
Times are stored as ISO 8601 strings in UTC
Only failed test results are stored (not passing tests)
The build_ingestion_tasks table shows whether test/timeline data is available for a build
Helix work items are fetched inline as part of the "tests" task (no separate helix task)
Known issues are refreshed every 15 minutes; closed issues are kept for 7 days before purging
Use LIKE '%pattern%' for fuzzy matching on test names, definition names, etc.
PR builds have source_branch like refs/pull/123/merge and pr_number set
Join builds to pull_requests on repository_name = repository AND pr_number = pr_number for PR details

User Display Rules

When displaying a build to the user include a link to the build
When displaying a Pr to the user include a link to the PR
Display times in the user's local timezone