agentfork-cli 0.1.0__tar.gz → 0.3.0__tar.gz

agentfork_cli-0.3.0/PKG-INFO

@@ -0,0 +1,196 @@
+Metadata-Version: 2.5
+Name: agentfork-cli
+Version: 0.3.0
+Summary: Fork your LLM agent at any step and sweep prompts, models or tool results to find what fixes it.
+Project-URL: Homepage, https://github.com/bimal1023/agentfork
+Project-URL: Issues, https://github.com/bimal1023/agentfork/issues
+Author: Bimal Kumal
+License-Expression: MIT
+License-File: LICENSE
+Keywords: agents,anthropic,debugging,langgraph,llm,openai,replay,testing
+Classifier: Development Status :: 3 - Alpha
+Classifier: Intended Audience :: Developers
+Classifier: Programming Language :: Python :: 3
+Classifier: Topic :: Software Development :: Debuggers
+Classifier: Topic :: Software Development :: Testing
+Requires-Python: >=3.9
+Requires-Dist: rich
+Requires-Dist: typer
+Provides-Extra: anthropic
+Requires-Dist: anthropic; extra == 'anthropic'
+Provides-Extra: dev
+Requires-Dist: anthropic; extra == 'dev'
+Requires-Dist: fastapi; extra == 'dev'
+Requires-Dist: httpx; extra == 'dev'
+Requires-Dist: langchain-openai; extra == 'dev'
+Requires-Dist: langgraph; extra == 'dev'
+Requires-Dist: openai; extra == 'dev'
+Requires-Dist: pytest; extra == 'dev'
+Requires-Dist: uvicorn; extra == 'dev'
+Provides-Extra: openai
+Requires-Dist: openai; extra == 'openai'
+Provides-Extra: ui
+Requires-Dist: fastapi; extra == 'ui'
+Requires-Dist: uvicorn; extra == 'ui'
+Description-Content-Type: text/markdown
+
+# agentfork
+
+[![tests](https://github.com/bimal1023/agentfork/actions/workflows/tests.yml/badge.svg)](https://github.com/bimal1023/agentfork/actions/workflows/tests.yml)
+
+**Fork your LLM agent at any step, then sweep prompts, models or tool results to find what actually fixes it.**
+
+![agentfork demo](https://raw.githubusercontent.com/bimal1023/agentfork/main/docs/demo.gif)
+
+When an agent misbehaves at step 3, the usual loop is: tweak the prompt, re-run everything, eyeball one output, repeat. That is slow, costs tokens on steps that didn't change, and one lucky run proves nothing. agentfork records every LLM and tool call, replays earlier steps from cache, and runs **many variants of one step, several times each**, so you get pass rates instead of anecdotes.
+
+Everything runs locally. No account, no server, no data leaves your machine.
+
+## The headline example
+
+A refund agent reads `"amount": 4999, "currency": "USD"` from an order API. The amount is in cents, and the agent refunds **$4,999** for a $49.99 laptop ([examples/refund_agent_cents.py](examples/refund_agent_cents.py)). Is the fix a better model or a better prompt?
+
+```text
+$ agentfork sweep 543da612 --step 3 --repeat 3 --expect "49.99" --model gpt-4o-mini,gpt-4o
+Estimated cost: about $0.0050
+
+  variant             ok    pass   tokens   time      cost   sample output
+  original            1/1   0/1       446   1.7s   $0.0001   I have processed a refund of $4,999.00...
+  model=gpt-4o-mini   3/3   0/3       446   1.4s   $0.0001   I have processed a refund of $4,999.00...
+  model=gpt-4o        3/3   0/3       394   1.2s   $0.0015   The full amount of $4,999.00 for order A123...
+```
+
+A bigger model costs 15x more per run and doesn't fix it. One sentence in the prompt does:
+
+```text
+$ agentfork sweep 543da612 --step 3 --repeat 3 --expect "49.99" \
+    --system "You are a support agent. Look up the order, then refund the full order amount in USD." \
+    --system "... Order amounts from the API are in cents; divide by 100 before refunding."
+Estimated cost: about $0.0006
+
+  variant     ok    pass   tokens   time      cost   sample output
+  original    1/1   0/1       446   1.7s   $0.0001   I have processed a refund of $4,999.00...
+  system #1   3/3   0/3       445   1.2s   $0.0001   I've successfully refunded you $4,999.00...
+  system #2   3/3   3/3       457   1.2s   $0.0001   I have processed a refund of $49.99...
+```
+
+Steps 1 and 2 replayed from cache in every run, so only the step that mattered was paid for, and the runs went 4 at a time.
+
+## Quickstart
+
+```bash
+pip install "agentfork-cli[openai]"      # or agentfork-cli[anthropic]
+```
+
+Add three things to your agent:
+
+```python
+from openai import OpenAI
+from agentfork import trace, tool, wrap_openai
+
+client = wrap_openai(OpenAI())      # records every LLM call (wrap_anthropic for Claude)
+
+@tool                               # records every tool call
+def lookup_order(order_id): ...
+
+@trace                              # one run = one call of your agent
+def run_agent(question): ...
+```
+
+Run your agent as usual. Each run is saved to `.agentfork/traces.db`.
+
+### LangGraph
+
+Wrap the chat model and put `@agentfork.tool` under LangChain's `@tool`. Forks and sweeps then work on the graph like on any other agent ([examples/langgraph_refund_agent.py](examples/langgraph_refund_agent.py)):
+
+```python
+import agentfork
+from langchain_core.tools import tool
+from langchain_openai import ChatOpenAI
+from langgraph.prebuilt import create_react_agent
+
+llm = agentfork.wrap_langchain(ChatOpenAI(model="gpt-4o-mini"))   # or ChatAnthropic
+
+@tool
+@agentfork.tool
+def lookup_order(order_id: str) -> str:
+    """Look up an order by id."""
+    ...
+
+graph = create_react_agent(llm, [lookup_order])
+
+@agentfork.trace
+def run_agent(question):
+    return graph.invoke({"messages": [("user", question)]})["messages"][-1].content
+```
+
+Streaming models (`streaming=True`) and the OpenAI Responses API aren't recorded yet.
+
+## Commands
+
+| Command | What it does |
+| --- | --- |
+| `agentfork list` | Recent runs with steps, tokens, duration and fork lineage |
+| `agentfork show <id> [--full]` | A run as a step tree, with optional full JSON |
+| `agentfork fork <id> --step N --output/--model/--system ...` | Re-run from step N with one change; earlier steps come from cache |
+| `agentfork sweep <id> --step N --model a,b --system ... --repeat K --expect TEXT [--parallel P]` | Try many variants of step N, K times each, P at a time (default 4), and compare pass rate, tokens, time and cost |
+| `agentfork diff <a> <b>` | Compare two runs step by step |
+| `agentfork save-test <id>` | Save a run as a pytest test that replays from cache in CI with no API key |
+| `agentfork ui [--port N]` | Open the local web UI to browse, fork, sweep and compare runs |
+
+## Web UI
+
+Prefer clicking to typing? The same fork, sweep and diff are available in a local web UI.
+
+```bash
+pip install "agentfork-cli[openai,ui]"
+agentfork ui        # opens http://127.0.0.1:8000
+```
+
+Start it from the folder you ran your agent in, since runs are saved to `.agentfork/` there. The server only listens on your own machine.
+
+1. **Find the step that went wrong.** Pick a run on the left. Its timeline shows each LLM and tool call in plain words, such as `Decided to call issue_refund(amount_usd=4999)`. Click a step to see its full input and output.
+2. **Try a change.** On that step, use another model, edit the system prompt, or fake a tool's result. Earlier steps replay from cache, and the new run appears under the original.
+3. **Test several.** Enter a few changes, how many times to run each, and text the final output must contain. You get a pass rate per change, with the original run as the baseline:
+
+   ```text
+   Change      Passed   Example output
+   Original    ○ 0/1    I have processed a refund of $4,999.00...
+   Change 1    ●●● 3/3  I have processed a refund of $49.99...
+   Change 2    ○○○ 0/3  I have processed a refund of $4,999.00...
+   ```
+
+4. **Compare.** Tick two runs to see them side by side, with the first step where they differ highlighted.
+
+## How it works
+
+1. **Record.** `@trace`, `@tool` and `wrap_openai`/`wrap_anthropic`/`wrap_langchain` capture each step's inputs, outputs, tokens and latency, plus a hash of its inputs.
+2. **Fork.** agentfork re-imports your agent and re-runs it. Steps before N are served from the cache, step N gets your change, and everything after runs live. A model change applies to every LLM call from step N on.
+3. **Sweep.** Runs fork for each variant `--repeat` times, `--parallel` at a time, and checks each final output for your `--expect` text. Before it starts, it estimates the cost from the tokens the original run used from step N on, priced per model.
+4. **Test.** `save-test` writes a JSON snapshot plus a pytest file. The test replays every step from cache and fails with a diff if any step's inputs, the step count, or the final output changes.
+
+## Costs
+
+Costs use list prices per model for OpenAI and Anthropic, built into [pricing.py](src/agentfork/pricing.py). Prices change, and a model without a price shows `-`. To add or correct one, put USD per 1M input and output tokens in `.agentfork/prices.json`:
+
+```json
+{"gpt-4o-mini": [0.15, 0.60], "my-finetune": [3.0, 12.0]}
+```
+
+## Related work
+
+agentfork overlaps with record/replay tools like [agent-rewind](https://pypi.org/project/agent-rewind/), which supports many more providers, and [agentstep](https://pypi.org/project/agentstep/), a LangGraph time-travel debugger with a web UI. agentfork's focus is the step after replay: sweeping many variants of one step, with repeats and pass rates, to choose a fix with evidence.
+
+## Roadmap
+
+- [x] Recorder, CLI viewer, fork, diff, snapshot tests
+- [x] OpenAI and Anthropic support
+- [x] Sweep with repeats and pass rates
+- [x] Local web UI
+- [x] Cost estimates per variant
+- [x] Parallel sweeps
+- [x] LangGraph integration
+
+## License
+
+MIT

agentfork_cli-0.3.0/README.md

@@ -0,0 +1,160 @@
+# agentfork
+
+[![tests](https://github.com/bimal1023/agentfork/actions/workflows/tests.yml/badge.svg)](https://github.com/bimal1023/agentfork/actions/workflows/tests.yml)
+
+**Fork your LLM agent at any step, then sweep prompts, models or tool results to find what actually fixes it.**
+
+![agentfork demo](https://raw.githubusercontent.com/bimal1023/agentfork/main/docs/demo.gif)
+
+When an agent misbehaves at step 3, the usual loop is: tweak the prompt, re-run everything, eyeball one output, repeat. That is slow, costs tokens on steps that didn't change, and one lucky run proves nothing. agentfork records every LLM and tool call, replays earlier steps from cache, and runs **many variants of one step, several times each**, so you get pass rates instead of anecdotes.
+
+Everything runs locally. No account, no server, no data leaves your machine.
+
+## The headline example
+
+A refund agent reads `"amount": 4999, "currency": "USD"` from an order API. The amount is in cents, and the agent refunds **$4,999** for a $49.99 laptop ([examples/refund_agent_cents.py](examples/refund_agent_cents.py)). Is the fix a better model or a better prompt?
+
+```text
+$ agentfork sweep 543da612 --step 3 --repeat 3 --expect "49.99" --model gpt-4o-mini,gpt-4o
+Estimated cost: about $0.0050
+
+  variant             ok    pass   tokens   time      cost   sample output
+  original            1/1   0/1       446   1.7s   $0.0001   I have processed a refund of $4,999.00...
+  model=gpt-4o-mini   3/3   0/3       446   1.4s   $0.0001   I have processed a refund of $4,999.00...
+  model=gpt-4o        3/3   0/3       394   1.2s   $0.0015   The full amount of $4,999.00 for order A123...
+```
+
+A bigger model costs 15x more per run and doesn't fix it. One sentence in the prompt does:
+
+```text
+$ agentfork sweep 543da612 --step 3 --repeat 3 --expect "49.99" \
+    --system "You are a support agent. Look up the order, then refund the full order amount in USD." \
+    --system "... Order amounts from the API are in cents; divide by 100 before refunding."
+Estimated cost: about $0.0006
+
+  variant     ok    pass   tokens   time      cost   sample output
+  original    1/1   0/1       446   1.7s   $0.0001   I have processed a refund of $4,999.00...
+  system #1   3/3   0/3       445   1.2s   $0.0001   I've successfully refunded you $4,999.00...
+  system #2   3/3   3/3       457   1.2s   $0.0001   I have processed a refund of $49.99...
+```
+
+Steps 1 and 2 replayed from cache in every run, so only the step that mattered was paid for, and the runs went 4 at a time.
+
+## Quickstart
+
+```bash
+pip install "agentfork-cli[openai]"      # or agentfork-cli[anthropic]
+```
+
+Add three things to your agent:
+
+```python
+from openai import OpenAI
+from agentfork import trace, tool, wrap_openai
+
+client = wrap_openai(OpenAI())      # records every LLM call (wrap_anthropic for Claude)
+
+@tool                               # records every tool call
+def lookup_order(order_id): ...
+
+@trace                              # one run = one call of your agent
+def run_agent(question): ...
+```
+
+Run your agent as usual. Each run is saved to `.agentfork/traces.db`.
+
+### LangGraph
+
+Wrap the chat model and put `@agentfork.tool` under LangChain's `@tool`. Forks and sweeps then work on the graph like on any other agent ([examples/langgraph_refund_agent.py](examples/langgraph_refund_agent.py)):
+
+```python
+import agentfork
+from langchain_core.tools import tool
+from langchain_openai import ChatOpenAI
+from langgraph.prebuilt import create_react_agent
+
+llm = agentfork.wrap_langchain(ChatOpenAI(model="gpt-4o-mini"))   # or ChatAnthropic
+
+@tool
+@agentfork.tool
+def lookup_order(order_id: str) -> str:
+    """Look up an order by id."""
+    ...
+
+graph = create_react_agent(llm, [lookup_order])
+
+@agentfork.trace
+def run_agent(question):
+    return graph.invoke({"messages": [("user", question)]})["messages"][-1].content
+```
+
+Streaming models (`streaming=True`) and the OpenAI Responses API aren't recorded yet.
+
+## Commands
+
+| Command | What it does |
+| --- | --- |
+| `agentfork list` | Recent runs with steps, tokens, duration and fork lineage |
+| `agentfork show <id> [--full]` | A run as a step tree, with optional full JSON |
+| `agentfork fork <id> --step N --output/--model/--system ...` | Re-run from step N with one change; earlier steps come from cache |
+| `agentfork sweep <id> --step N --model a,b --system ... --repeat K --expect TEXT [--parallel P]` | Try many variants of step N, K times each, P at a time (default 4), and compare pass rate, tokens, time and cost |
+| `agentfork diff <a> <b>` | Compare two runs step by step |
+| `agentfork save-test <id>` | Save a run as a pytest test that replays from cache in CI with no API key |
+| `agentfork ui [--port N]` | Open the local web UI to browse, fork, sweep and compare runs |
+
+## Web UI
+
+Prefer clicking to typing? The same fork, sweep and diff are available in a local web UI.
+
+```bash
+pip install "agentfork-cli[openai,ui]"
+agentfork ui        # opens http://127.0.0.1:8000
+```
+
+Start it from the folder you ran your agent in, since runs are saved to `.agentfork/` there. The server only listens on your own machine.
+
+1. **Find the step that went wrong.** Pick a run on the left. Its timeline shows each LLM and tool call in plain words, such as `Decided to call issue_refund(amount_usd=4999)`. Click a step to see its full input and output.
+2. **Try a change.** On that step, use another model, edit the system prompt, or fake a tool's result. Earlier steps replay from cache, and the new run appears under the original.
+3. **Test several.** Enter a few changes, how many times to run each, and text the final output must contain. You get a pass rate per change, with the original run as the baseline:
+
+   ```text
+   Change      Passed   Example output
+   Original    ○ 0/1    I have processed a refund of $4,999.00...
+   Change 1    ●●● 3/3  I have processed a refund of $49.99...
+   Change 2    ○○○ 0/3  I have processed a refund of $4,999.00...
+   ```
+
+4. **Compare.** Tick two runs to see them side by side, with the first step where they differ highlighted.
+
+## How it works
+
+1. **Record.** `@trace`, `@tool` and `wrap_openai`/`wrap_anthropic`/`wrap_langchain` capture each step's inputs, outputs, tokens and latency, plus a hash of its inputs.
+2. **Fork.** agentfork re-imports your agent and re-runs it. Steps before N are served from the cache, step N gets your change, and everything after runs live. A model change applies to every LLM call from step N on.
+3. **Sweep.** Runs fork for each variant `--repeat` times, `--parallel` at a time, and checks each final output for your `--expect` text. Before it starts, it estimates the cost from the tokens the original run used from step N on, priced per model.
+4. **Test.** `save-test` writes a JSON snapshot plus a pytest file. The test replays every step from cache and fails with a diff if any step's inputs, the step count, or the final output changes.
+
+## Costs
+
+Costs use list prices per model for OpenAI and Anthropic, built into [pricing.py](src/agentfork/pricing.py). Prices change, and a model without a price shows `-`. To add or correct one, put USD per 1M input and output tokens in `.agentfork/prices.json`:
+
+```json
+{"gpt-4o-mini": [0.15, 0.60], "my-finetune": [3.0, 12.0]}
+```
+
+## Related work
+
+agentfork overlaps with record/replay tools like [agent-rewind](https://pypi.org/project/agent-rewind/), which supports many more providers, and [agentstep](https://pypi.org/project/agentstep/), a LangGraph time-travel debugger with a web UI. agentfork's focus is the step after replay: sweeping many variants of one step, with repeats and pass rates, to choose a fix with evidence.
+
+## Roadmap
+
+- [x] Recorder, CLI viewer, fork, diff, snapshot tests
+- [x] OpenAI and Anthropic support
+- [x] Sweep with repeats and pass rates
+- [x] Local web UI
+- [x] Cost estimates per variant
+- [x] Parallel sweeps
+- [x] LangGraph integration
+
+## License
+
+MIT

agentfork_cli-0.3.0/examples/langgraph_refund_agent.py

@@ -0,0 +1,33 @@
+import json
+from langchain_core.tools import tool
+from langchain_openai import ChatOpenAI
+from langgraph.prebuilt import create_react_agent
+import agentfork
+
+# The refund agent from refund_agent_cents.py, built with LangGraph.
+# wrap_langchain records the model's calls; @agentfork.tool (under @tool) records the tools.
+llm = agentfork.wrap_langchain(ChatOpenAI(model="gpt-4o-mini"))
+
+SYSTEM = "You are a support agent. Look up the order, then refund the full order amount in USD."
+ORDERS = {"A123": {"order_id": "A123", "item": "laptop", "amount": 4999, "currency": "USD", "status": "delivered"}}
+
+@tool
+@agentfork.tool
+def lookup_order(order_id: str) -> str:
+    """Look up an order by id."""
+    return json.dumps(ORDERS.get(order_id, {"error": "not found"}))
+
+@tool
+@agentfork.tool
+def issue_refund(order_id: str, amount_usd: float) -> str:
+    """Refund an order. amount_usd is in US dollars."""
+    return f"Refunded ${amount_usd:,.2f} for order {order_id}"
+
+graph = create_react_agent(llm, [lookup_order, issue_refund], prompt=SYSTEM)
+
+@agentfork.trace
+def run_agent(question):
+    return graph.invoke({"messages": [("user", question)]})["messages"][-1].content
+
+if __name__ == "__main__":
+    print(run_agent("My headphones from order A123 arrived broken. Please refund me."))

{agentfork_cli-0.1.0 → agentfork_cli-0.3.0}/pyproject.toml

@@ -1,12 +1,12 @@
 [project]
 name = "agentfork-cli"
-version = "0.1.0"
+version = "0.3.0"
 description = "Fork your LLM agent at any step and sweep prompts, models or tool results to find what fixes it."
 readme = "README.md"
 requires-python = ">=3.9"
 license = "MIT"
 authors = [{ name = "Bimal Kumal" }]
-keywords = ["llm", "agents", "debugging", "replay", "testing", "openai", "anthropic"]
+keywords = ["llm", "agents", "debugging", "replay", "testing", "openai", "anthropic", "langgraph"]
 classifiers = [
     "Development Status :: 3 - Alpha",
     "Intended Audience :: Developers",
@@ -19,7 +19,8 @@ dependencies = ["typer", "rich"]
 [project.optional-dependencies]
 openai = ["openai"]
 anthropic = ["anthropic"]
-dev = ["pytest", "openai", "anthropic"]
+ui = ["fastapi", "uvicorn"]
+dev = ["pytest", "openai", "anthropic", "fastapi", "uvicorn", "httpx", "langgraph", "langchain-openai"]
 
 [project.urls]
 Homepage = "https://github.com/bimal1023/agentfork"
@@ -34,3 +35,6 @@ build-backend = "hatchling.build"
 
 [tool.hatch.build.targets.sdist]
 exclude = ["docs/", ".github/", "tests/agentfork/"]
+
+[tool.hatch.build.targets.wheel]
+packages = ["src/agentfork"]

agentfork_cli-0.3.0/src/agentfork/__init__.py

@@ -0,0 +1,3 @@
+from .recorder import trace, tool, wrap_anthropic, wrap_langchain, wrap_openai
+
+__all__ = ["trace", "tool", "wrap_openai", "wrap_anthropic", "wrap_langchain"]

{agentfork_cli-0.1.0 → agentfork_cli-0.3.0}/src/agentfork/cli.py

@@ -3,12 +3,13 @@ from datetime import datetime
 from typing import List, Optional
 
 import typer
+from rich import box
 from rich.console import Console
 from rich.syntax import Syntax
 from rich.table import Table
 from rich.tree import Tree
 
-from . import replay, store
+from . import pricing, replay, store
 
 app = typer.Typer(help="agentfork: record, inspect and replay LLM agent runs.")
 console = Console()
@@ -84,10 +85,13 @@ def show(run_id: str = typer.Argument(..., help="Run id or its first few charact
     color = STATUS_COLOR.get(run["status"], "white")
     fork = (f" [dim]fork of {run['parent_run_id'][:8]} at step {run['fork_step']}[/dim]"
             if run.get("parent_run_id") else "")
+    steps = store.get_steps(run["id"])
+    cost = pricing.steps_cost([s for s in steps if not s["cached"]])
+    cost = f" [dim]{pricing.fmt(cost)}[/dim]" if cost is not None else ""
     root = Tree(f"[bold]{run['name']}[/bold] {run['id'][:8]} "
-                f"[{color}]{run['status']}[/{color}] {_duration(run)}{fork}")
+                f"[{color}]{run['status']}[/{color}] {_duration(run)}{cost}{fork}")
     nodes = {}
-    for s in store.get_steps(run["id"]):
+    for s in steps:
         node = nodes.get(s["parent_id"], root).add(_label(s))
         if full:
             node.add(Syntax(json.dumps(_load(s["input_json"]), indent=2), "json", word_wrap=True))
@@ -195,6 +199,13 @@ def _make_labels(variants):
             labels[id(v)] = f"{key} #{counts[key]}"
     return labels
 
+def _clip(text, n):
+    return text if len(text) <= n else text[:n - 1] + "…"
+
+def _avg_cost(rows):
+    costs = [r["cost"] for r in rows]
+    return "-" if None in costs else pricing.fmt(sum(costs) / len(costs))
+
 @app.command()
 def sweep(run_id: str = typer.Argument(..., help="Run to sweep."),
           step: int = typer.Option(..., "--step", help="Step to vary (see `show`)."),
@@ -202,7 +213,8 @@ def sweep(run_id: str = typer.Argument(..., help="Run to sweep."),
           system: Optional[List[str]] = typer.Option(None, "--system", help="System prompt to try; repeat for more."),
           output: Optional[List[str]] = typer.Option(None, "--output", help="Tool result to try; repeat for more."),
           repeat: int = typer.Option(1, "--repeat", help="Runs per variant (LLMs vary run to run)."),
-          expect: Optional[str] = typer.Option(None, "--expect", help="Text a good final output contains.")):
+          expect: Optional[str] = typer.Option(None, "--expect", help="Text a good final output contains."),
+          parallel: int = typer.Option(4, "--parallel", help="How many runs to make at the same time.")):
     """Fork one step across several models, prompts or tool results and compare outcomes."""
     variants = [{"model": m.strip()} for m in (model or "").split(",") if m.strip()]
     variants += [{"system": s} for s in (system or [])]
@@ -213,7 +225,17 @@ def sweep(run_id: str = typer.Argument(..., help="Run to sweep."),
     run = _resolve(run_id)
     labels = _make_labels(variants)
     console.print(f"[bold]Sweeping step {step} of {run['id'][:8]}: "
-                  f"{len(variants)} variants x {repeat} = {len(variants) * repeat} runs[/bold]")
+                  f"{len(variants)} variants x {repeat} = {len(variants) * repeat} runs, "
+                  f"{min(parallel, len(variants) * repeat)} at a time[/bold]")
+    try:
+        est = replay.estimate(run, step, variants, repeat)
+    except ValueError:
+        est = None
+    if est and est["total"] is not None:
+        console.print(f"[dim]Estimated cost: about {pricing.fmt(est['total'])}[/dim]")
+    elif est and est["unpriced"]:
+        console.print(f"[dim]Estimated cost unknown: no price for {', '.join(est['unpriced'])} "
+                      f"(add it to .agentfork/prices.json)[/dim]")
 
     def progress(r):
         mark = "[green]ok[/green]" if r["status"] == "ok" else "[red]error[/red]"
@@ -221,7 +243,8 @@ def sweep(run_id: str = typer.Argument(..., help="Run to sweep."),
         console.print(f"  {mark}{verdict} {labels[id(r['variant'])]} [dim]{r['tokens']} tok[/dim]")
 
     try:
-        results = replay.sweep(run, step, variants, repeat=repeat, expect=expect, on_result=progress)
+        results = replay.sweep(run, step, variants, repeat=repeat, expect=expect,
+                               on_result=progress, parallel=parallel)
     except ValueError as e:
         console.print(f"[red]{e}[/red]")
         raise typer.Exit(1)
@@ -229,31 +252,50 @@ def sweep(run_id: str = typer.Argument(..., help="Run to sweep."),
     groups = {"original": [replay.baseline(run, step, expect)]}
     for r in results:
         groups.setdefault(labels[id(r["variant"])], []).append(r)
-    table = Table(title=f"Sweep of step {step} ({run['id'][:8]})", show_lines=True)
+    table = Table(title=f"Sweep of step {step} ({run['id'][:8]})", box=box.SIMPLE_HEAD,
+                  caption="tokens, time and cost are averages per run", caption_justify="left")
     table.add_column("variant", no_wrap=True)
-    table.add_column("runs", no_wrap=True, min_width=4)
-    table.add_column("ok", no_wrap=True, min_width=3)
+    table.add_column("ok", no_wrap=True)
     if expect:
-        table.add_column("pass", no_wrap=True, min_width=4)
-    table.add_column("avg tokens", no_wrap=True, min_width=6)
-    table.add_column("avg time", no_wrap=True, min_width=6)
-    table.add_column("sample output", max_width=36, overflow="ellipsis", no_wrap=True)
-    table.add_column("run ids", no_wrap=True)
+        table.add_column("pass", no_wrap=True)
+    table.add_column("tokens", no_wrap=True, justify="right")
+    table.add_column("time", no_wrap=True, justify="right")
+    table.add_column("cost", no_wrap=True, justify="right")
+    table.add_column("sample output", max_width=40)
+    table.add_column("runs", no_wrap=True)
     for label, rs in groups.items():
         n = len(rs)
         ids = [(r["run_id"] or "-")[:8] for r in rs]
-        row = [label, str(n), f"{sum(r['status'] == 'ok' for r in rs)}/{n}"]
+        row = [label, f"{sum(r['status'] == 'ok' for r in rs)}/{n}"]
         if expect:
             passed = sum(bool(r["passed"]) for r in rs)
             color = "green" if passed == n else ("yellow" if passed else "red")
             row.append(f"[{color}]{passed}/{n}[/{color}]")
         row += [f"{sum(r['tokens'] for r in rs) / n:.0f}",
                 f"{sum(r['latency_ms'] for r in rs) / n / 1000:.1f}s",
-                str(rs[0]["result"]),
-                ", ".join(ids[:2]) + (f" +{len(ids) - 2}" if len(ids) > 2 else "")]
+                _avg_cost(rs),
+                _clip(str(rs[0]["result"]), 60 if console.width >= 100 else 30),
+                ids[0] + (f" +{n - 1}" if n > 1 else "")]
         table.add_row(*row)
     console.print(table)
     for v in variants:
         key, value = next(iter(v.items()))
         if key != "model":
             console.print(f"[bold]{labels[id(v)]}[/bold]: {value}")
+
+@app.command()
+def ui(port: int = typer.Option(8000, "--port", help="Port for the local web UI."),
+       no_browser: bool = typer.Option(False, "--no-browser", help="Don't open a browser.")):
+    """Open the local web UI: runs, timelines, fork, sweep and compare."""
+    try:
+        import uvicorn
+        from .server import app as web_app
+    except ImportError:
+        console.print('[red]The UI needs extra packages: pip install "agentfork-cli[ui]"[/red]')
+        raise typer.Exit(1)
+    url = f"http://127.0.0.1:{port}"
+    console.print(f"[green]agentfork UI running at {url}[/green]  (Ctrl+C to stop)")
+    if not no_browser:
+        import threading, webbrowser
+        threading.Timer(1.0, lambda: webbrowser.open(url)).start()
+    uvicorn.run(web_app, host="127.0.0.1", port=port, log_level="warning")

agentfork_cli-0.3.0/src/agentfork/pricing.py

@@ -0,0 +1,70 @@
+"""Model prices for cost estimates, in USD per 1M tokens: (input, output).
+
+Standard-tier list prices, checked October 2026. Prices change, so add or
+override models in .agentfork/prices.json, for example:
+
+    {"gpt-4o-mini": [0.15, 0.60], "my-finetune": [3.0, 12.0]}
+"""
+import json, re, warnings
+from . import store
+
+PRICES = {
+    # OpenAI
+    "gpt-5": (1.25, 10.00), "gpt-5-mini": (0.25, 2.00), "gpt-5-nano": (0.05, 0.40),
+    "gpt-4.1": (2.00, 8.00), "gpt-4.1-mini": (0.40, 1.60), "gpt-4.1-nano": (0.10, 0.40),
+    "gpt-4o": (2.50, 10.00), "gpt-4o-mini": (0.15, 0.60),
+    "o3": (2.00, 8.00), "o3-mini": (1.10, 4.40), "o4-mini": (1.10, 4.40),
+    # Anthropic
+    "claude-fable-5-1": (10.00, 50.00), "claude-fable-5": (10.00, 50.00),
+    "claude-opus-5-5": (4.00, 20.00), "claude-opus-5": (5.00, 25.00),
+    "claude-opus-4-8": (5.00, 25.00), "claude-opus-4-7": (5.00, 25.00),
+    "claude-opus-4-6": (5.00, 25.00), "claude-opus-4-5": (5.00, 25.00),
+    "claude-opus-4-1": (15.00, 75.00), "claude-opus-4": (15.00, 75.00),
+    "claude-sonnet-5-5": (2.00, 10.00), "claude-sonnet-5": (2.00, 10.00),
+    "claude-sonnet-4-6": (3.00, 15.00), "claude-sonnet-4-5": (3.00, 15.00),
+    "claude-sonnet-4": (3.00, 15.00), "claude-3-7-sonnet": (3.00, 15.00),
+    "claude-haiku-5-5": (0.10, 0.50), "claude-haiku-4-5": (1.00, 5.00),
+    "claude-3-5-haiku": (0.80, 4.00),
+}
+
+# Dated snapshots and aliases cost the same as the base model: gpt-4o-mini-2024-07-18, claude-3-5-haiku-latest.
+_SUFFIX = re.compile(r"-(\d{8}|\d{4}-\d{2}-\d{2}|latest)$")
+
+def _user_prices():
+    path = store.db_path().parent / "prices.json"
+    try:
+        return {k.lower(): tuple(v) for k, v in json.loads(path.read_text()).items()}
+    except FileNotFoundError:
+        return {}
+    except (ValueError, TypeError, AttributeError) as e:
+        warnings.warn(f"agentfork: ignoring {path}: {e}")
+        return {}
+
+def price(model):
+    """(input, output) USD per 1M tokens, or None if the model is unknown."""
+    if not model:
+        return None
+    table = {**PRICES, **_user_prices()}
+    name = model.lower()
+    return table.get(name) or table.get(_SUFFIX.sub("", name))
+
+def steps_cost(steps, model=None):
+    """Total USD for the LLM steps, priced at `model` if given. None if a model or token count is unknown."""
+    total = 0.0
+    for s in steps:
+        if s["type"] != "llm":
+            continue
+        p = price(model or s["name"])
+        if p is None or s["tokens_in"] is None:
+            return None
+        total += (s["tokens_in"] * p[0] + (s["tokens_out"] or 0) * p[1]) / 1e6
+    return total
+
+def unpriced(steps, model=None):
+    """Models among the LLM steps that have no price."""
+    return sorted({model or s["name"] for s in steps if s["type"] == "llm" and price(model or s["name"]) is None})
+
+def fmt(cost):
+    if cost is None:
+        return "-"
+    return f"${cost:.4f}" if cost < 0.01 else f"${cost:.2f}"