jsat 0.4.0

Codebase intelligence shell and SDK — index once, query forever, works with Claude, GPT, Ollama, and any AI

JSAT — JaySoft AI Tools

Codebase intelligence for AI sessions — index once, query forever, works with any AI.

---

What is JSAT?

Every AI session starts with the same problem: you spend the first ten minutes re-explaining your architecture, re-pasting function signatures, and re-describing how services talk to each other. JSAT solves this by building a persistent graph of your codebase once — functions, classes, files, services, API endpoints, database tables, Kafka topics, and every relationship between them — and making that context instantly available to any AI you use.

JSAT works as a CLI, a Python SDK, and an MCP server that plugs directly into Claude Code. If Claude Code CLI is installed, JSAT uses it automatically with no API key required. For everything else — Anthropic API, OpenAI, Gemini, Ollama, LM Studio — one command switches the provider.

Long-running tools (jsat__crack, jsat__query, jsat__short, prompt rewriting) stream live progress notifications back to Claude Code so you see what's happening in real time — no more blank screens during 30-second operations.

---

Quick Start

pip install jsat

# Index your project
cd your-project/
jsat index .

# Open your AI tool with JSAT pre-loaded (auto-connects on first use)
jsat claude      # Claude Code
jsat codex       # OpenAI Codex CLI
jsat cursor      # Cursor IDE
jsat windsurf    # Windsurf
jsat gemini      # Google Gemini CLI
jsat zed         # Zed editor
jsat bob         # Bob Shell

# War room discussion (new!)
jsat crack "redesign payment retry system"

# Shortest possible answer (new!)
jsat short "what does process_refund do"

Inside any connected tool you can use JSAT commands:

Claude Code slash commands:

/jsat query what does the payment service do?
/jsat blast-radius src/payment/refund.py
/jsat security
/jsat incident "500 errors spiking on checkout"
/jsat prompt-rewrite fix logger in ValidateVPAHandler.post

Continue.dev custom commands (same commands, /jsat-* prefix)

All tools have all JSAT MCP tools the AI can call automatically — no slash commands needed.

---

Installation

JSAT ships as a minimal core with optional extras. Install only what you need.

Extra	What's added	Approx. size	When to use
(none) / core	tree-sitter parsers, SQLite graph, CLI	~80 MB	Starting point for any setup
local	Ollama client	+small	Local models via Ollama
standard	Semgrep, OpenAPI/AsyncAPI validator, more language parsers (Java, Ruby, Rust)	+medium	Security reviews, API contract checks
team	Neo4j, Qdrant, Redis, Graphiti (includes standard)	+large	Shared graph across a team
anthropic	Anthropic Python SDK	+small	Claude API (key required)
openai	OpenAI Python SDK	+small	GPT-4o, GPT-4o-mini (key required)
ci	PyGitHub, SARIF tools (includes standard)	+small	CI/CD pipelines, GitHub Actions
all	Everything above	+large	Full feature set

pip install jsat                   # core only
pip install 'jsat[local]'          # + Ollama
pip install 'jsat[standard]'       # + security analysis, OpenAPI validation
pip install 'jsat[team]'           # + Neo4j, Qdrant, Redis
pip install 'jsat[anthropic]'      # + Claude API SDK
pip install 'jsat[openai]'         # + OpenAI SDK
pip install 'jsat[all]'            # everything

---

AI Providers

JSAT auto-detects available providers at startup and picks the best one in priority order:

1. Claude Code CLI — detected via which claude; no API key, no extra SDK, full tool calling
2. Bob Shell CLI — detected via which bob; no API key, IBM AI assistant with multiple modes
3. Anthropic API — if ANTHROPIC_API_KEY is set and jsat[anthropic] is installed
4. OpenAI — if OPENAI_API_KEY is set and jsat[openai] is installed
5. Ollama — if ollama serve is running at localhost:11434
6. LM Studio — if an OpenAI-compatible server is running at localhost:1234
7. No AI — tools that don't need AI (indexing, blast radius, export) still work

Check what's available

jsat ai status        # shows all providers, which is active, and switch commands

Switch providers

jsat ai use ollama                        # free, local, no key
jsat ai use ollama --model qwen2.5-coder:7b
jsat ai use anthropic                     # needs ANTHROPIC_API_KEY
jsat ai use openai --model gpt-4o-mini    # needs OPENAI_API_KEY
jsat ai use claude_cli                    # Claude Code CLI (no key, uses claude binary)
jsat ai use lmstudio                      # any OpenAI-compat server at localhost:1234
jsat ai test                              # verify the configured provider works

# Apply globally (all projects on this machine):
jsat ai use claude_cli --global

Switch inside the JSAT shell

switch claude    → Claude Code CLI (no key) or Claude API
switch bob       → Bob Shell (no key)
switch gpt       → GPT-4o
switch ollama    → local Ollama
switch haiku     → Claude Haiku
switch phi       → phi3:mini (fast, low RAM)
switch lmstudio  → LM Studio

---

AI Tool Integration

JSAT works as an MCP server with any AI tool that supports the Model Context Protocol. One command wires it in — the tool picks up all JSAT MCP tools automatically.

Connect

# Recommended: one-time global setup (works in every project)
jsat connect claude --global               # Claude Code — all sessions
jsat connect codex --global               # OpenAI Codex CLI — all sessions
jsat connect bob --global                 # Bob Shell — all sessions

# Per-project (this repo only)
jsat connect claude                        # Claude Code
jsat connect codex                         # OpenAI Codex CLI
jsat connect bob                           # Bob Shell (+ /jsat-* slash commands)
jsat connect cursor                        # Cursor
jsat connect windsurf                      # Windsurf (Codeium)
jsat connect continue                      # Continue.dev
jsat connect zed                           # Zed editor
jsat connect gemini                        # Google Gemini CLI

jsat connect list                          # show every active connection

Restart the AI tool after connecting. JSAT's MCP tools are immediately available.

Files written per tool

Each connect command writes both an MCP config and a guidance file so the AI knows what JSAT tools exist and when to use them — without being asked.

Tool	MCP config	Guidance file	Guidance format
Claude Code (project)	.claude/settings.json	.claude/commands/jsat-*.md (31 files)	Slash commands
Claude Code (global)	~/.claude/settings.json	~/.claude/commands/jsat-*.md	Slash commands
Codex (project)	.codex/config.json	.codex/instructions.md	Agent instructions
Codex (global)	~/.codex/config.json	~/.codex/instructions.md	Agent instructions
Bob Shell (project)	.bob/settings.json	BOB.md + .bob/commands/jsat-*.md	Slash commands
Bob Shell (global)	~/.bob/settings.json	BOB.md + ~/.bob/commands/jsat-*.md	Slash commands
Cursor	~/.cursor/mcp.json	—	—
Windsurf	~/.codeium/windsurf/mcp_config.json	.windsurfrules	Rules file
Continue	~/.continue/config.json	/jsat-* custom commands	Slash commands
Zed	~/.config/zed/settings.json	.zed/JSAT.md	Project context
Gemini CLI	~/.gemini/settings.json	GEMINI.md	Project instructions

Pass --global (claude/codex/bob) or check the tool's docs for global scope on others.

Pass --no-instructions to skip writing the guidance file (MCP only).

/jsat dispatcher

jsat connect claude installs a single /jsat command rather than 34 individual /jsat-* commands. All skills are accessible as subcommands:

/jsat help               # list all 34 subcommands
/jsat query <question>   # answer codebase questions (Discuss→Verify pipeline)
/jsat crack <task>       # multi-agent war room with artifact carry-forward
/jsat aw <task>          # workflow advisor
/jsat lazy <task>        # reuse-first planning
/jsat smart <question>   # terse compressed answers

Skill files are bundled inside the JSAT package at jsat/commands/ and sourced directly when jsat connect claude runs.

Disconnect

jsat disconnect claude                     # Claude Code project scope
jsat disconnect claude --scope all         # Claude Code everywhere
jsat disconnect codex                      # Codex
jsat disconnect cursor                     # Cursor
jsat disconnect windsurf                   # Windsurf
jsat disconnect continue                   # Continue
jsat disconnect zed                        # Zed
jsat disconnect gemini                     # Gemini CLI
jsat disconnect all                        # every tool at once

Claude Code — slash commands (34 total)

jsat connect claude installs a single /jsat dispatcher with 34 subcommands, organized by category:

Graph exploration

Command	What it does
/jsat query <question>	Natural language query over the indexed graph
/jsat find-function <name>	Look up a function — file, params, return type, complexity
/jsat find-class <name>	Look up a class — file, bases, method count
/jsat list-services	List all indexed services
/jsat list-endpoints	List all API endpoints with method, route, auth
/jsat trace <symbol>	Trace a call chain from a symbol
/jsat index [path]	Rebuild the codebase graph (incremental)
/jsat status	Node/edge counts
/jsat doctor	Full system health check

Impact & safety

Command	What it does
/jsat blast-radius <file or symbol>	Downstream impact grouped by severity
/jsat security [path]	OWASP scan — Critical and High first
/jsat migration <file>	DB migration safety — lock type, duration estimate
/jsat contract <diff>	API contract compatibility check

Code quality

Command	What it does
/jsat review <diff>	Multi-model parallel code review
/jsat test-gaps [path]	Find untested paths, generate tests
/jsat coverage [path]	Behavioral coverage estimate

Knowledge & investigation

Command	What it does
/jsat knowledge <query>	Search the knowledge base
/jsat knowledge-add <text>	Add an ADR / runbook / decision
/jsat runbook <target>	Generate an incident runbook
/jsat incident <description>	Root-cause hypotheses ranked by confidence
/jsat recent [path]	Recent changes in an area

Prompt & token tools

Command	What it does
/jsat prompt <query>	Optimize a prompt through the full pipeline
/jsat prompt-diff <query>	Show raw input vs what the AI actually received
/jsat tokens <text>	Count tokens; compress to fit context limit
/jsat token-budget <text>	Check budget against the active model's context window

IThinking

Command	What it does
/jsat ithinking <task>	Full IThinking: plan → assumptions → decompose → confirm
/jsat think <task>	Quick shortcut — think before any task
/jsat reflect <outcome>	Record what was done (phase 6 log)

Analysis & planning

Command	What it does
/jsat crack <task>	Multi-agent war room with artifact carry-forward
/jsat short <question>	Minimum-word answer (≤3 sentences)
/jsat smart <question>	Terse compressed answer, filler stripped
/jsat lazy <task>	Reuse-first planning — 5-rung ladder before suggesting new code
/jsat aw <task>	Workflow advisor — classify task, run optimal tool sequence

Open Claude with JSAT context pre-loaded

jsat claude

---

JSAT Crack — Multi-Agent War Room

Run a complex engineering decision past a panel of six specialist AI agents that argue, challenge, and respond to each other — like a real architecture meeting.

jsat crack "redesign payment retry system"
jsat crack --roles architect,security "migrate users table to UUID"
jsat crack --rounds 2 --file design.md "sync vs async webhooks"

Agents (all run in parallel per round, then respond to each other):

Agent	Focus
🏛 architect	System design, scalability, patterns
🔒 security	Threat model, auth, idempotency
⚙️ implementer	Current code analysis, effort estimate
🧪 tester	Edge cases, coverage gaps
😈 skeptic	Devil's advocate — challenges everything
🎯 moderator	Synthesises consensus and action plan

Phased mode (v0.4.0+): By default, each agent runs in its own phase (N=6) so you see results after every agent instead of waiting for all 6. Each agent receives all prior findings as context — the architect's conclusions inform the security analysis, and the skeptic specifically challenges the architect's and implementer's proposals.

Flag	Effect
--phases N	Override phase count (2–6, default 6)
--single	One-shot: all 6 agents at once (may timeout)

jsat crack "redesign the payment retry system"
jsat crack --phases 4 "add idempotency to the charge endpoint"  
jsat crack --single "should we use Redis or Postgres for sessions?"

Output: A Markdown document saved to .jsat/crack/<slug>.md with each round's discussion and a final synthesis:

✅ Agreed:        Use exponential backoff with tenacity
⚠️ Disputed:      Redis vs in-process lock for idempotency
❓ Open questions: What's the SLA for retry exhaustion?
🎯 Action:        1. Extract retry logic, 2. Add idempotency key, 3. Write tests

In Claude Code: /jsat crack redesign the payment retry system

In the JSAT shell: crack should we use async or sync for webhook processing

MCP tool: jsat__crack

Live progress — while the war room runs you see real-time updates in Claude Code:

⚡ Loading codebase context…
⚡ Round 1/3: Opening statements…
⚡ Round 1/3: Moderator synthesising…
⚡ Round 2/3: Cross-examination…
⚡ Round 3/3: Consensus…
⚡ Writing discussion document…
[result]

---

JSAT Short — Minimum-Word Answers

Get the shortest possible correct answer to any question.

jsat short "what does process_refund do"
jsat short --one-line "is PaymentService.process async"
jsat short --words 20 "explain the retry logic"

In Claude Code: /jsat short what does process_refund do

In the JSAT shell: short is the checkout flow async

---

Prompt Optimizer

JSAT optimizes every query through a two-phase pipeline before sending to the AI.

Phase 1 — Offline pipeline (always runs, zero LLM calls)

jsat prompt "improve the retry logic"              # inspect optimized prompt
jsat prompt --send "improve the retry logic"       # optimize + send
jsat prompt --diff "improve the retry logic"       # see raw vs optimized side by side
jsat prompt --send --format code --ai claude "write a test for refund()"

Stage	What it does
Classify	Keyword-match task type: code_gen / refactor / debug / test / security / …
Context	BFS graph traversal — injects relevant function signatures and call chains
Constraints	KB lookup — injects project ADRs and coding standards (top-3 only)
Few-shot	kNN over prompt history — injects the most similar past examples
Format	Provider-aware: XML for Claude, Markdown for GPT, plain for Ollama
Compress	Token pruning when prompt exceeds 4000 tokens

Phase 2 — LLM rewriting (optional, activated with a flag)

After the offline pipeline structures the prompt, 1–3 specialist LLM agents rewrite the task description from different angles, then the best result is selected by a coverage + specificity scorer.

# 1 agent — fastest, rewrites for clarity and precision
jsat prompt --rewrite "fix logger in this branch"

# 3 agents in parallel — picks the best rewrite
jsat prompt --agents "fix logger in this branch"

# Combine with --send to optimize + rewrite + send in one step
jsat prompt --agents --send "fix logger in ValidateVPAHandler.post"

The 3 LLM agents:

Agent	Temperature	Focus
rewrite	0.2	Replaces vague words with specific identifiers from context
context_expand	0.3	Fills missing technical detail (function names, error messages, paths)
constraint_harden	0.1	Makes success criteria measurable ("ensure X returns Y when Z")

Agents run in parallel. Winner is chosen by: coverage × 0.45 + specificity × 0.40 + efficiency × 0.15.

Discuss → Verify pipeline (v0.4.0+): Before optimizing, the skill classifies the query type and selects the right primary tool — structural questions use jsat__trace_call_chain, lookup questions use jsat__get_function, security questions use jsat__security_review, etc. After executing, Phase 5 spot-checks concrete claims from the answers against the graph, marking each as ✅ verified or ⚠️ unverified before synthesis.

Example output with --agents --verbose:

┌─────────────────────┬──────────────────────────────┐
│ Task type           │ debug                        │
│ Context nodes       │ 3                            │
│ Tokens before       │ 6                            │
│ Tokens after        │ 847                          │
│ Rewrite agents run  │ 3                            │
│ Winner              │ context_expand               │
│ Rewrite time        │ 1843ms                       │
└─────────────────────┴──────────────────────────────┘

✦ 6→847 tokens | Task: debug | 3 agents → context_expand won

In the shell — every message is auto-optimized through Phase 1:

jsat [Claude Code (CLI)]> improve the retry logic

✦ Optimized refactor | 6→847 tokens (35% saved) | 3 ctx nodes | opt show to see diff

Claude: Here's the improved retry using tenacity...

Shell commands:

opt on        # enable auto-optimization (default)
opt off       # disable for the current session
opt show      # show raw input vs full optimized prompt side by side
opt history   # browse past optimization diffs

MCP tools:

• jsat__prompt_optimize — offline pipeline only
• jsat__prompt_rewrite — offline + 1 LLM rewrite agent
• jsat__prompt_multi_agent — offline + up to 3 parallel LLM agents

Claude Code slash command:

/jsat prompt-rewrite fix the logger missing extra= dict in ValidateVPAHandler.post

Disconnect or remove

jsat disconnect claude --scope all            # Claude Code everywhere
jsat disconnect all                           # every connected tool at once
jsat remove                                   # remove all JSAT artifacts from this repo

---

JSAT Smart — Terse Mode

Get compressed, fragment-based answers with filler stripped. Preserves code, function names, file paths, and data byte-for-byte.

Three compression levels:

Level	Flag	Reduction	What it does
lite	--lite	~30%	Strips filler phrases only ("In order to", "It's worth noting", etc.)
full	(default)	~55%	Fragments + no explanatory preamble
ultra	--ultra	~70%	One bullet per fact, ≤8 words each

jsat smart "what does the payment service do?"
jsat smart --ultra "what does process_refund return?"
jsat smart --lite "explain the checkout flow"

Use as a fast fallback when /jsat query times out on large contexts.

---

JSAT Lazy — Reuse-First Planning

Before writing new code, runs a 5-rung reuse ladder against the indexed codebase graph. Stops at the first match.

Rung	What it checks	Tool used
1	Exact function/class already exists?	jsat__get_function / jsat__get_class
2	Similar pattern in codebase?	jsat__query
3	Existing service handles this domain?	jsat__list_services
4	Existing endpoint already exposes this?	jsat__list_endpoints
5	Nothing found → minimum viable code	(suggest only)

# Check before building new retry logic
/jsat lazy add exponential backoff to the payment service

# Scan a diff for code that reimplements existing things
/jsat lazy --audit src/payments/retry.py

# Check if a proposed implementation duplicates existing code
/jsat lazy --review "def process_refund(order_id, amount)..."

---

JSAT Aw — Workflow Advisor

Classifies your task type and runs the optimal JSAT tool sequence end-to-end — no more guessing which tool to use or in what order.

Task type	Recommended workflow
feature	lazy → find-function → blast-radius → crack → test-gaps
bugfix	recent → incident → find-function → blast-radius
security	security → blast-radius --severity breaking → crack --phases 3 → knowledge-add
understand	smart → trace → find-function → query
incident	incident → recent → blast-radius → runbook
refactor	lazy → blast-radius → test-gaps → crack → review
review	review → blast-radius --severity breaking → test-gaps --untested

# Classify automatically and run the full workflow
/jsat aw add idempotency keys to the payment mutation endpoint

# Skip classification, force a specific workflow type
/jsat aw --type security src/auth/

# Show the recommended workflow without running it
/jsat aw --dry investigate the checkout 500 errors from this morning

---

CLI Reference

Core commands

Command	Description
jsat index [path]	Build or update the codebase graph (incremental, parallel)
jsat index . --force	Full re-index — ignore incremental manifest
jsat index . --watch	Re-index on file change (requires brew install entr)
jsat index . --languages python,go	Index specific languages only
jsat shell	Start the interactive JSAT REPL
jsat claude	Open Claude Code with JSAT MCP tools loaded
jsat codex	Open Codex CLI with JSAT pre-loaded
jsat cursor	Open Cursor IDE with JSAT pre-loaded
jsat windsurf	Open Windsurf with JSAT pre-loaded
jsat gemini	Open Gemini CLI with JSAT pre-loaded
jsat zed	Open Zed editor with JSAT pre-loaded
jsat gpt	Open a GPT session with JSAT tools
jsat ollama [--model llama3.2]	Open a local Ollama session
jsat crack <task>	Multi-agent war room discussion (6 AI specialists)
jsat crack --roles architect,security <task>	War room with specific roles only
jsat crack --rounds 2 --file out.md <task>	2 rounds, save output to file
jsat short <question>	Get the shortest possible correct answer
jsat short --one-line <question>	Exactly one-sentence answer
jsat prompt <query>	Print the optimized prompt (inspect without sending)
jsat prompt --send <query>	Optimize prompt and send to the configured AI
jsat prompt --rewrite <query>	Run 1 LLM rewrite agent after offline pipeline
jsat prompt --agents <query>	Run 3 parallel LLM rewrite agents, pick best
jsat prompt --diff <query>	Show raw input vs optimized prompt side by side
jsat doctor	System health check (graph, AI, services, connected tools)
jsat doctor --json	Health check as raw JSON
jsat version	Print JSAT version

Configuration

Command	Description
jsat init	Generate .jsat/config.yaml for this repo (default: solo profile)
jsat init --global	Generate ~/.jsat/config.yaml — applies to all projects
jsat init --profile team	Team profile (Neo4j, Qdrant, Redis, Claude API)
jsat init --profile ci	CI profile (SQLite, no AI, JSON logs)
jsat init --profile raspberry-pi	Low-RAM profile (SQLite, phi3:mini, batch size 8)

AI provider management

Command	Description
jsat ai status	Show all providers: available, active, free/paid
jsat ai use <provider>	Configure a provider and write to .jsat/config.yaml
jsat ai use <provider> --global	Configure provider globally in ~/.jsat/config.yaml
jsat ai use ollama --model phi3:mini	Use a specific Ollama model
jsat ai test	Send a test prompt and verify the provider works
jsat ai models	List models available from the configured provider

Claude Code integration

Command	Description
jsat connect claude	Wire JSAT into Claude Code (project scope) + install 31 slash commands
jsat connect claude --global	Wire JSAT into Claude Code globally (all projects)
jsat connect claude --no-skills	MCP only — skip slash command installation
jsat connect codex	Wire JSAT into OpenAI Codex CLI (project scope)
jsat connect codex --global	Wire JSAT into Codex globally
jsat connect bob	Wire JSAT into Bob Shell (project scope)
jsat connect bob --global	Wire JSAT into Bob Shell globally
jsat connect cursor	Wire JSAT into Cursor
jsat connect windsurf	Wire JSAT into Windsurf
jsat connect continue	Wire JSAT into Continue.dev
jsat connect zed	Wire JSAT into Zed editor
jsat connect gemini	Wire JSAT into Google Gemini CLI
jsat connect list	Show all active JSAT MCP connections
jsat disconnect <tool>	Remove JSAT from a specific tool
jsat disconnect all	Remove JSAT from every tool at once

Token analysis

Command	Description
jsat tokens "text"	Count tokens in inline text
jsat tokens --file README.md	Count tokens in a file
jsat tokens --file ctx.py --model gpt-4o	Show budget bar vs model limit
jsat tokens --file ctx.py --compress	Compress and show savings
jsat tokens --target 4000 --compress	Compress to explicit token ceiling
cat file.py | jsat tokens --model claude-cli	Pipe stdin

Export and import

Command	Description
jsat export backup.jsat.zip	Export the current index as a portable zip
jsat export backup.jsat.zip -z 9	Export with maximum compression
jsat import backup.jsat.zip	Restore an index from an exported archive

Skills

Command	Description
jsat skills list	List installed JSAT skill manifests
jsat skills run <name>	Run a named skill with optional key=val args
jsat ci-setup	Write a GitHub Actions workflow for JSAT
jsat ci-setup --provider gitlab	Write a GitLab CI pipeline for JSAT

---

Python SDK

from jsat import JSAT

# Instantiate — auto-detects AI provider, loads config
js = JSAT(repo=".")

# Build the graph — parallel parsing, incremental by default
result = js.index()
print(f"Indexed {result.nodes_indexed} nodes in {result.duration_ms}ms")
print(f"Workers: {result.parallel_workers} | Incremental: {result.incremental}")
print(f"Skipped: {result.files_skipped} unchanged | Resolved: {result.resolved_edges} edges")
if result.complexity_hotspots:
    print("Hotspots:", [(h["name"], h["complexity"]) for h in result.complexity_hotspots])

# Natural language query over the graph
result = js.query("what calls the refund endpoint?")
print(result.answer)

# Trace blast radius of a change
report = js.blast_radius("src/payment/refund.py")
for item in report.impacts:
    print(f"{item.severity:10s}  {item.node_id}")

# Security analysis (requires jsat[standard])
sec = js.security_review(path="src/")
for finding in sec.findings:
    print(f"{finding.severity}: {finding.title} — {finding.file}:{finding.line}")

# Incident investigation
incident = js.investigate_incident("500 errors on checkout", since="24h")
for h in incident.hypotheses:
    print(f"[{h.score:.0%}] {h.title}")

# Export the index for sharing or CI caching
manifest = js.export("snapshot.jsat.zip")
print(f"Exported {manifest.size_mb:.1f} MB")

# Restore from an export
js2 = JSAT.from_import("snapshot.jsat.zip")

# Token analysis (offline, no LLM)
count = js.token_count("explain the payment service")
report = js.token_compress(large_prompt, model="gpt-4o")
print(f"Saved {report.savings_pct:.1f}% via: {report.strategies_applied}")
budget = js.token_budget(my_context, "claude-cli")
print(f"{budget['budget_pct']:.2f}% of context used ({budget['status']})")

# Switch AI provider mid-session
js.switch_ai("ollama", model="qwen2.5-coder:7b")
js.switch_ai("anthropic")
js.switch_ai("gpt", model="gpt-4o-mini")

# Health check
health = js.doctor()
print(health["profile"], health["graph"]["backend"])

---

Tools Overview (20 tools)

#	Tool	Description
0	JSAT Shell	Interactive REPL — run any tool directly, switch AI mid-session, no AI required
1	Directory Indexer	Parallel tree-sitter parsing (4–8× faster), true incremental mode, rich metadata (parameters, return types, decorators, docstrings, complexity), symbol resolution, inheritance/raises edges
2	Test Intelligence Helper	Finds test gaps, maps behaviors to coverage, generates unit/integration/contract tests
3	Feature Helper	Answers "how do I add X?" using graph context — finds relevant files and patterns
4	Blast Radius Analyzer	BFS over the graph to trace downstream impact; classifies edges as breaking/degraded/warning/safe
5	API Contract Validator	Diffs OpenAPI/AsyncAPI specs, classifies breaking changes, scores backward compatibility (0–100)
6	Security Review Agent	OWASP pattern scan, auth coverage gaps, hardcoded secret detection, dependency CVE lookup
7	Incident Investigation Helper	Correlates an incident description against recent commits and graph topology; ranks root-cause hypotheses
8	Migration Safety Validator	Validates migration files, estimates lock duration, generates zero-downtime migration plans
9	Multi-Model Code Review (true parallel dispatch)	Dispatches a diff to multiple AI models simultaneously via ThreadPoolExecutor; surfaces only bugs confirmed by two or more models
10	Knowledge Base Builder	Persistent searchable store of architectural decisions, runbooks, and tribal knowledge
11	Multi-Agent Orchestrator	Decomposes a task and runs specialized sub-agents (understanding, generation, review, test, security, docs)
12	Export / Import System	Portable zip snapshots of the full graph — share between machines, cache in CI, restore in seconds
13	Python SDK	Programmatic access to every tool via from jsat import JSAT
14	IThinking Meta-Cognitive Layer	Structured seven-phase reasoning: clarify, plan, context, assumptions, execute, reflect — with human approval gates
15	Token Optimizer	Offline token analysis and multi-strategy compression (whitespace, stopphrase, dedup, import collapse, recency pin)
16	JSAT Crack	Multi-agent war room — 6 specialists discuss complex decisions in rounds, responding to each other; moderator synthesises consensus
17	JSAT Short	Minimum-word answers — prepends brevity constraint so AI responds in ≤3 sentences or less
18	JSAT Smart	Terse compression mode — fragment-based answers with filler stripped, code preserved
19	JSAT Lazy	Reuse-first planning — 5-rung ladder against the graph before suggesting new code
20	JSAT Aw	Workflow advisor — classifies task type, runs optimal JSAT tool sequence end-to-end

Multi-Model Review

Tool 9 dispatches the diff to all configured models in parallel, collects findings, and surfaces only those confirmed by two or more models. Configure the model list and timeout in .jsat/config.yaml:

review:
  models:
    - {provider: claude_cli, model: claude-sonnet-4-6}
    - {provider: ollama, model: qwen2.5-coder:7b}
  parallel_timeout_seconds: 90
  min_confidence: medium

• parallel_timeout_seconds — per-review wall-clock deadline; any model that exceeds this is skipped and its absence is logged.
• min_confidence — minimum agreement level to surface a finding: low (any model), medium (2+ models), high (all models).

---

Graph Schema

JSAT indexes these node types and relationship edges:

Nodes: Function, Class, File, Service, Endpoint, Table, Topic, KnowledgeEntry

Node properties (v0.2.0+):

Property	On	Example
name, file, language, line_start, line_end, line	Function, Class	"PaymentService.refund"
parameters	Function	[{"name":"order_id","type":"str"}]
return_type	Function	"bool", "list[Payment]"
decorators	Function, Class	["staticmethod","login_required"]
docstring	Function, Class	first line, max 200 chars
complexity	Function	cyclomatic (1 + branch count)
loc	Function	line_end - line_start + 1
bases	Class	["BaseModel","Serializable"]
method_count	Class	number of methods in class body

Edges:

Edge	Meaning
CALLS	Function A calls function B (resolved to node ID post-parse)
IMPORTS	File A imports module B
INHERITS	Class inherits from parent (all 6 languages)
IMPLEMENTS	Class implements interface/trait (Java, Go, Rust)
RAISES	Function can raise exception type (Python)
READS_FROM	Code reads from a table or topic
WRITES_TO	Code writes to a table or topic
PRODUCES	Service produces a Kafka message
CONSUMES	Service consumes a Kafka topic
DEPENDS_ON	Service depends on another service

---

Configuration

Data storage

By default JSAT stores runtime state (graph, cache, vectors, prompt history) in ~/.jsat/<hash12>/ — a global per-repo directory that never appears inside your git working tree. No .gitignore entry needed.

Priority	Location	When
1	$JSAT_DATA_DIR	env var set (CI, Docker)
2	{repo}/.jsat/	already exists (backward compat)
3	~/.jsat/<sha1_12>/	default

The config file (.jsat/config.yaml) is separate and optional — it holds project-specific settings, not runtime data.

One-time global setup

jsat init --global --profile solo      # write ~/.jsat/config.yaml
jsat ai use claude_cli --global        # set AI provider globally
jsat connect claude --global           # install to ~/.claude/ for all projects

After this, every project on the machine has JSAT available with no per-repo steps.

Per-repo config

jsat init                        # write .jsat/config.yaml (solo profile)
jsat init --profile team         # team profile
jsat init --profile ci           # CI/CD profile

Key settings in .jsat/config.yaml:

graph:
  backend: sqlite          # sqlite (default) | neo4j (team profile)
  path: .jsat/graph/graph.db

embeddings:
  provider: local          # local | openai | none
  model: nomic-embed-code

ai:
  provider: ollama         # ollama | anthropic | openai | openai_compat | claude_cli | none
  model: llama3.2
  base_url: null           # for openai_compat (LM Studio, Gemini, etc.)

cache:
  backend: memory          # memory | disk | redis

indexer:
  languages: [python, javascript, go, java, ruby, rust]
  exclude_patterns: ["**/node_modules/**", "**/.git/**", "**/dist/**"]
  max_file_size_kb: 500

ithinking:
  enabled: true
  mode: interactive        # interactive | silent
  gate_level: medium       # low | medium | high

Profiles at a glance

Profile	Graph	AI	Cache	Use case
solo	SQLite	Ollama / llama3.2	Memory	Individual developer, no external services
team	Neo4j	Claude API	Redis	Shared graph, team-wide knowledge base
ci	SQLite	None	Memory	GitHub Actions, no API keys, JSON logs
raspberry-pi	SQLite	Ollama / phi3:mini	Disk	Low-RAM devices, batch size 8

Config search order

JSAT finds its config file by checking these locations in order (first found wins):

1. Explicit path passed to JSAT(config=...) or --config flag
2. $JSAT_CONFIG environment variable
3. {repo}/.jsat/config.yaml (repo-local canonical)
4. {repo}/.jsat.yaml (legacy)
5. ./.jsat/config.yaml (CWD)
6. ~/.jsat/config.yaml ← global user config (jsat init --global)
7. ~/.config/jsat/config.yaml
8. /etc/jsat/config.yaml

---

Supported Languages

Language	Parser	Notes
Python	tree-sitter-python	Core (always available)
JavaScript / TypeScript	tree-sitter-javascript	Core
Go	tree-sitter-go	Core
Java	tree-sitter-java	Requires jsat[standard]
Ruby	tree-sitter-ruby	Requires jsat[standard]
Rust	tree-sitter-rust	Requires jsat[standard]

---

Contributing

Contributions are welcome. Please open an issue or pull request on GitHub.

• Repository: github.com/iamjpsonkar/JaySoft-AI_Tools
• Bug reports: open an issue on GitHub
• Author: Jay Prakash Sonkar — [e-mail]
• License: MIT

---

License

MIT License. Copyright (c) Jay Prakash Sonkar.