| name | run-parseltongue-1-7-2 |
| description | Install and run parseltongue 1.7.2 in an isolated user-space run folder, verify the actual bound server URL, and query a codebase through the HTTP API for orientation, entity lookup, forward and backward dependency tracing, blast radius, architecture scans, incremental reindexing, and higher-order graph workflows such as meet-in-the-middle debugging, safe refactoring, seam discovery, boundary audits, cycle breaking, migration choreography, and derived control-flow or data-flow investigations. Use when Codex must pin to version 1.7.2, needs graph-based code understanding, or must refresh an existing run folder without version drift. |
Run Parseltongue 1.7.2
Use this skill to install parseltongue 1.7.2, create a fresh run folder, index a target codebase, launch the query server, and run initial API queries.
Default retention policy: clean run folders older than 1 day before each new setup.
Default storage location: ${CODEX_HOME:-$HOME/.codex}/tool-runs/parseltongue-1-7-2.
Ground Truth
Treat these as non-obvious but binding:
- Do not assume
pt01-folder-to-cozodb-streamer --db ... will honor the requested DB path.
- Trust the
DB_URI and WORKSPACE_PATH recorded in run.config, not an assumed db/codegraph.db path.
- Verify both health and watcher status before assuming live auto-reindex works.
- Treat server logs and changed-symbol queries as the strongest proof of incremental reindex; watcher counters can lag or stay at zero.
- Keep the Parseltongue run folder outside any directory that may later be zipped or uploaded.
- Prefer the default user-space run home unless the user explicitly wants a different
--base-dir.
- Do not assume
http://localhost:7777 is the live server after setup or refresh; Parseltongue can auto-select the next available port.
- Prefer
SERVER_URL in run.config when present. If it is missing, read the actual bound server URL from <run-folder>/logs/server.log.
- Treat
api-reference-documentation-help as the live source of truth for endpoint names and parameters.
Enforce Version Policy
Follow these rules on every use:
- Pin to
v1.7.2 only.
- Verify
parseltongue --version contains 1.7.2 before indexing or querying.
- Create a new timestamped run folder every setup.
Default name pattern:
parseltongue-1.7.2-<codebase-slug>-<timestamp>.
- Keep binary, workspace DB, logs, and query artifacts inside that run folder.
- Refuse to proceed if version check fails.
- Clean folders older than 1 day by default during setup.
Code Reading Preference
Prefer Parseltongue queries to read and understand code.
Use direct file reads (rg, cat, editor open) only if relevant entities or relationships cannot be found through Parseltongue, or coverage/endpoint diagnostics show the needed area is not available.
Flow Modeling Guardrail
Treat control-flow and data-flow conclusions from Parseltongue as graph-derived approximations, not perfect runtime truth.
These APIs are strongest for dependency topology, caller pressure, downstream spread, cluster boundaries, architectural reach, and change risk.
When a flow question depends on branch conditions, exact value propagation, dynamic dispatch, reflection, generated code, or low-coverage files, use Parseltongue to narrow the search space first and then confirm with direct code reads.
Run Workflow
- Run installer and setup script:
./scripts/setup_parseltongue_1_7_2.sh /path/to/codebase
- Capture the printed run folder path, workspace path, DB URI, and actual server URL. If the requested port is busy, setup may bind to the next available port.
- Load
<run-folder>/run.config and prefer SERVER_URL, DB_URI, WORKSPACE_PATH, and CODEBASE_PATH from there.
- Verify health, watcher status, and the new-codebase orientation queries:
curl -fsS "${SERVER_URL}/server-health-check-status"
curl -fsS "${SERVER_URL}/file-watcher-status-check"
curl -fsS "${SERVER_URL}/codebase-statistics-overview-summary"
curl -fsS "${SERVER_URL}/circular-dependency-detection-scan"
curl -fsS "${SERVER_URL}/complexity-hotspots-ranking-view?top=10"
curl -fsS "${SERVER_URL}/semantic-cluster-grouping-list"
curl -fsS "${SERVER_URL}/ingestion-coverage-folder-report?depth=2"
curl -fsS "${SERVER_URL}/folder-structure-discovery-tree"
curl -fsS "${SERVER_URL}/api-reference-documentation-help"
- If watcher health is false or stale, inspect
<run-folder>/logs/server.log and <run-folder>/logs/index.log before assuming the index is usable.
If watcher status looks ambiguous, confirm with a small code edit and a code-entities-search-fuzzy query for the new symbol.
- Run focused analysis queries or a bidirectional investigation workflow:
curl -fsS "${SERVER_URL}/code-entities-search-fuzzy?q=main"
curl -fsS "${SERVER_URL}/code-entity-detail-view?key=ENTITY_KEY"
curl -fsS "${SERVER_URL}/reverse-callers-query-graph?entity=ENTITY_KEY"
curl -fsS "${SERVER_URL}/forward-callees-query-graph?entity=ENTITY_KEY"
curl -fsS "${SERVER_URL}/blast-radius-impact-analysis?entity=ENTITY_KEY&hops=2"
curl -fsS "${SERVER_URL}/smart-context-token-budget?focus=ENTITY_KEY&tokens=4000"
- Stop server when done:
./scripts/stop_parseltongue_server.sh /path/to/run-folder
Bidirectional Workflows
- Trace backward from a symptom:
Search the broken function, read its code, inspect
reverse-callers-query-graph, then quantify depth with blast-radius-impact-analysis. Use this for bug hunting and trigger discovery.
- Trace forward from a candidate change:
Read the entity, inspect
forward-callees-query-graph, then quantify downstream impact with blast-radius-impact-analysis and pull the surrounding code with smart-context-token-budget. Use this for implementation planning and dependency drilling.
- Quantify refactor risk:
Check direct callers, blast radius with
hops=3, circular dependencies, and then request a larger smart-context budget such as tokens=8000. Treat 0-5 direct callers as low risk, 6-15 as medium, and more than 15 as high. If blast radius exceeds 100 at hops=2, assume you are touching critical infrastructure.
- Load advanced recipes when the user wants creative graph workflows such as meet-in-the-middle debugging, cycle-breaking, deletion safety, migration choreography, boundary-leak detection, or seam discovery:
Read references/parseltongue_1_7_2_bidirectional_workflows.md.
- Load derived flow recipes when the user wants approximated control flow, data flow, lineage, ingress-to-egress tracing, transformer ladders, orchestrator detection, or boundary-crossing analysis:
Read references/parseltongue_1_7_2_flow_patterns.md.
Script Parameters
Use setup script options as needed:
./scripts/setup_parseltongue_1_7_2.sh /path/to/codebase \
--base-dir "${CODEX_HOME:-$HOME/.codex}/tool-runs/parseltongue-1-7-2" \
--port 7777 \
--hops 2 \
--cleanup-days 1
--base-dir: Parent directory that receives a new timestamped run folder. Default: ${CODEX_HOME:-$HOME/.codex}/tool-runs/parseltongue-1-7-2.
--port: HTTP server port.
--hops: Blast radius hop count for initial query capture.
--cleanup-days: Remove run folders older than this many days before setup (default 1).
After setup, use <run-folder>/run.config as the canonical source for:
SERVER_URL
DB_URI
WORKSPACE_PATH
CODEBASE_PATH
PORT
REQUESTED_PORT
Reindex Existing Run
Reindex an existing run folder without re-downloading:
./scripts/reindex_parseltongue_1_7_2.sh /path/to/run-folder
Optional overrides:
./scripts/reindex_parseltongue_1_7_2.sh /path/to/run-folder \
--codebase-path /path/to/codebase \
--port 7777
If the user says "refresh", "refresh index", or "reindex", run this reindex script.
Use refresh when:
- watcher health is false
- live file edits are not appearing in queries
- the workspace DB is stale or suspect
- the server URL or bound port may have changed after restart and must be re-read from
run.config or server.log
Global Shortcuts
Use global shell wrappers (available from any directory):
pt172-setup /path/to/codebase
pt172-refresh /path/to/run-folder
pt172-stop /path/to/run-folder
Query Guidance
Read references/parseltongue_1_7_2_endpoints.md for:
- the 5-query orientation pass
- endpoint selection by question
- backward bug-hunting flow
- forward dependency-tracing flow
- safe-refactoring risk checks
- diagnostics and incremental reindex examples
Read references/parseltongue_1_7_2_bidirectional_workflows.md for:
- meet-in-the-middle hypothesis testing
- symptom-to-system corridor tracing
- migration choreography
- narrow-waist seam discovery
- boundary-leak detection
- cycle-breaker scouting
- deletion-safety classification
Read references/parseltongue_1_7_2_flow_patterns.md for:
- control-flow approximations from the dependency graph
- data-flow and lineage approximations from the dependency graph
- ingress-to-egress tracing
- orchestrator versus transformer classification
- model or payload gravity analysis