dicklesworthstone__coding_agent_session_search/tests/fixtures

Test Fixtures

This directory contains real fixture data for integration and E2E tests. Fixtures follow the project's no-mock policy — see TESTING.md.

Directory Overview

Directory	Purpose	Test Scenarios
aider/	Aider chat history format	Markdown history parsing
amp/	Amp thread format	JSON thread parsing
claude_code_real/	Real Claude Code sessions	JSONL session parsing
claude_project/	Claude project structure	Project directory discovery
cli_contract/	CLI API contract fixtures	Version/capability stability
cline/	Cline task format	Multi-file task parsing
codex_real/	Real Codex CLI sessions	JSONL session parsing
connectors/	Connector manifest	Fixture provenance tracking
gemini/	Gemini session format	Hash-based directory parsing
html_export/	HTML export test data	Edge cases, real sessions
install/	Installation artifacts	Installer script testing
models/	ONNX model fixtures	Embedding/reranker loading
multi_source/	Multi-machine sources	Source sync testing
opencode_json/	OpenCode JSON format	Message/part/session parsing
pages_verify/	Page verification	Security/validity checks
pi_agent/	Pi Agent sessions	Session format parsing
search_demo_data/	Pre-indexed search data	Search/query testing
sources/	Sources subsystem	Probe result fixtures

---

Fixture Details

aider/

Scenario: Parsing Aider's markdown-based chat history format.

File	Description
.aider.chat.history.md	Minimal Aider session with /add command and refactoring request

Tested by: src/connectors/aider.rs

---

amp/

Scenario: Parsing Amp's JSON thread format.

File	Description
thread-001.json	Sample thread with user/assistant messages and timestamps

Tested by: src/connectors/amp.rs

---

claude_code_real/

Scenario: Parsing real Claude Code JSONL session files.

Path	Description
projects/-test-project/agent-test123.jsonl	Real session with matrix completion discussion

Tested by: src/connectors/claude_code.rs, tests/e2e_search_index.rs

---

claude_project/

Scenario: Discovering Claude Code project directories.

Path	Description
projectA/	Sample project directory structure

Tested by: src/connectors/claude_code.rs (project discovery)

---

cli_contract/

Scenario: CLI API contract stability and backwards compatibility.

File	Description
api_version.json	API version response fixture
capabilities.json	Feature capabilities (connectors, limits, features)
introspect.json	Full introspection response for CLI contract testing

Tested by: tests/e2e_cli_contract.rs

Use case: Verifies CLI JSON output matches expected schema across versions.

---

cline/

Scenario: Parsing Cline's multi-file task format.

Path	Description
task1/api_conversation_history.json	API conversation history
task1/task_metadata.json	Task metadata
task1/ui_messages.json	UI-facing messages

Tested by: src/connectors/cline.rs

---

codex_real/

Scenario: Parsing real Codex CLI JSONL sessions.

Path	Description
sessions/2025/11/25/rollout-test.jsonl	Codex CLI session fixture

Tested by: src/connectors/codex.rs, tests/e2e_search_index.rs

---

connectors/

Scenario: Provenance tracking for all connector fixtures.

File	Description
MANIFEST.json	Central manifest with SHA256 checksums, capture dates, redaction policies

Use case: Validates fixture integrity and documents data provenance.

---

gemini/

Scenario: Parsing Gemini's hash-based directory structure.

Path	Description
hash123/	Sample Gemini session directory

Tested by: src/connectors/gemini.rs

---

html_export/

Scenario: HTML export functionality with various edge cases.

Subdirectory	Description
edge_cases/	Boundary conditions for HTML generation
malformed/	Invalid/corrupted input handling
real_sessions/	Real session data from 11 different agents

Edge case files:

File	Scenario
all_message_types.jsonl	Every message type in one session
empty_session.jsonl	Zero messages
large_session.jsonl	357KB session (performance test)
single_message.jsonl	Minimal valid session
unicode_heavy.jsonl	CJK, emoji, RTL text rendering

Real session coverage:

Agent	Session
Aider	aider_bugfix.jsonl
Amp	amp_data_pipeline.jsonl
ChatGPT	chatgpt_react_help.jsonl
Claude Code	claude_code_auth_fix.jsonl
Cline	cline_vscode_setup.jsonl
Codex	codex_api_design.jsonl
Cursor	cursor_refactoring.jsonl
Factory	factory_code_generation.jsonl
Gemini	gemini_debugging.jsonl
OpenCode	opencode_rust_cli.jsonl
Pi Agent	pi_agent_personal_assistant.jsonl

Tested by: tests/e2e_html_export.rs, src/html_export/

---

install/

Scenario: Testing the installation script with mock artifacts.

File	Description
coding-agent-search	Mock Linux binary
coding-agent-search.exe	Mock Windows binary
*.tar.gz, *.zip	Mock release archives
*.sha256	Checksum files for integrity verification

Tested by: tests/e2e_install_easy.rs

---

models/

Scenario: ONNX model loading for semantic search and reranking.

See models/README.md for full documentation.

Subdirectory	Description
Root files	Minimal valid ONNX model for unit tests
xenova-paraphrase-minilm-l3-v2-int8/	Real embedding model (~17MB)
xenova-ms-marco-minilm-l6-v2-int8/	Real reranker model (~22MB)

Tested by: src/search/embedder.rs, src/search/reranker.rs, tests/e2e_semantic.rs

---

multi_source/

Scenario: Multi-machine source synchronization.

Subdirectory	Description
local/	Local machine sessions
remote_laptop/	Remote laptop sessions (different paths)
remote_workstation/	Remote workstation sessions

Tested by: tests/e2e_multi_source.rs, tests/e2e_sources.rs

---

opencode_json/

Scenario: Parsing OpenCode's JSON format at multiple levels.

Subdirectory	Description
message/	Individual message fixtures
part/	Message part fixtures
session/proj1/	Complete session fixtures

Tested by: src/connectors/opencode.rs

---

pages_verify/

Scenario: Static page verification for security and correctness.

Subdirectory	Test Case
valid/site/	Correctly structured exported site
missing_required/	Missing required files (should fail)
missing_required_no_viewer/	Missing viewer.html (should fail)
secret_leak/	Accidental credential exposure detection
unencrypted/	Unencrypted export verification

Tested by: tests/e2e_pages.rs

---

pi_agent/

Scenario: Parsing Pi Agent session format.

Path	Description
sessions/	Pi Agent session fixtures

Tested by: src/connectors/pi_agent.rs

---

search_demo_data/

Scenario: Pre-indexed search data for query testing.

File/Dir	Description
agent_search.db	Pre-populated SQLite database
index/v1/	Frozen legacy lexical index fixture (migration testing)
index/v*/	Locally generated current-schema indexes are ignored and should be rebuilt by tests
watch_state.json	File watcher state fixture

Tested by: tests/e2e_search_index.rs, src/search/query.rs

---

sources/

Scenario: Sources subsystem probe fixtures.

See sources/probe/README.md for full documentation.

Subdirectory	Description
probe/	HostProbeResult JSON fixtures for various states

Tested by: src/sources/probe.rs, tests/e2e_sources.rs

---

Adding New Fixtures

1. Use real data — Capture from actual agent sessions when possible
2. Redact sensitive info — Anonymize usernames, paths, credentials
3. Add to MANIFEST — Update connectors/MANIFEST.json with SHA256
4. Document the scenario — Update this README and add fixture-specific README if complex
5. Link to tests — Note which test files exercise the fixture

Fixture Loading Helpers

Common fixture loading utilities are in tests/fixture_helpers.rs:

use fixture_helpers::{
    load_fixture,                    // Load fixture file as string
    copy_fixture,                    // Copy fixture to temp location
    embedder_fixture_dir,            // Path to embedding model fixtures
    reranker_fixture_dir,            // Path to reranker model fixtures
    verify_model_fixture_checksums,  // Verify ONNX model integrity
    setup_connector_test,            // Create temp test environment
    create_project_dir,              // Create project within test env
    write_session_file,              // Write session file to test env
};

No-Mock Policy

Per TESTING.md, this project avoids mocks in favor of real fixtures:

• Connector tests use real session files from each agent
• Database tests use real SQLite with known data
• HTTP tests use embedded servers with fixture responses
• Model tests use real (quantized) ONNX models

This approach provides higher confidence that tests reflect production behavior.