Rollback

This guide assumes knowledge of what double-texting is, which you can learn about in the double-texting conceptual guide.

The guide covers the rollback option for double texting, which interrupts the prior run of the graph and starts a new one with the double-text. This option is very similar to the interrupt option, but in this case the first run is completely deleted from the database and cannot be restarted. Below is a quick example of using the rollback option.

Setup

First, we will define a quick helper function for printing out JS and CURL model outputs (you can skip this if using Python):

JavascriptCURL

function prettyPrint(m) {
  const padded = " " + m['type'] + " ";
  const sepLen = Math.floor((80 - padded.length) / 2);
  const sep = "=".repeat(sepLen);
  const secondSep = sep + (padded.length % 2 ? "=" : "");

  console.log(`${sep}${padded}${secondSep}`);
  console.log("\n\n");
  console.log(m.content);
}

# PLACE THIS IN A FILE CALLED pretty_print.sh
pretty_print() {
  local type="$1"
  local content="$2"
  local padded=" $type "
  local total_width=80
  local sep_len=$(( (total_width - ${#padded}) / 2 ))
  local sep=$(printf '=%.0s' $(eval "echo {1.."${sep_len}"}"))
  local second_sep=$sep
  if (( (total_width - ${#padded}) % 2 )); then
    second_sep="${second_sep}="
  fi

  echo "${sep}${padded}${second_sep}"
  echo
  echo "$content"
}

Now, let's import our required packages and instantiate our client, assistant, and thread.

PythonJavascriptCURL

import asyncio

import httpx
from langchain_core.messages import convert_to_messages
from langgraph_sdk import get_client

client = get_client(url=<DEPLOYMENT_URL>)
# Using the graph deployed with the name "agent"
assistant_id = "agent"
thread = await client.threads.create()

import { Client } from "@langchain/langgraph-sdk";

const client = new Client({ apiUrl: <DEPLOYMENT_URL> });
// Using the graph deployed with the name "agent"
const assistantId = "agent";
const thread = await client.threads.create();

curl --request POST \
  --url <DEPLOYMENT_URL>/threads \
  --header 'Content-Type: application/json' \
  --data '{}'

Create runs

Now let's run a thread with the multitask parameter set to "rollback":

PythonJavascriptCURL

# the first run will be rolled back
rolled_back_run = await client.runs.create(
    thread["thread_id"],
    assistant_id,
    input={"messages": [{"role": "user", "content": "what's the weather in sf?"}]},
)
run = await client.runs.create(
    thread["thread_id"],
    assistant_id,
    input={"messages": [{"role": "user", "content": "what's the weather in nyc?"}]},
    multitask_strategy="rollback",
)
# wait until the second run completes
await client.runs.join(thread["thread_id"], run["run_id"])

// the first run will be interrupted
let rolledBackRun = await client.runs.create(
  thread["thread_id"],
  assistantId,
  { input: { messages: [{ role: "human", content: "what's the weather in sf?" }] } }
);

let run = await client.runs.create(
  thread["thread_id"],
  assistant_id,
  { 
    input: { messages: [{ role: "human", content: "what's the weather in nyc?" }] },
    multitaskStrategy: "rollback" 
  }
);

// wait until the second run completes
await client.runs.join(thread["thread_id"], run["run_id"]);

curl --request POST \
--url <DEPLOY<ENT_URL>>/threads/<THREAD_ID>/runs \
--header 'Content-Type: application/json' \
--data "{
  \"assistant_id\": \"agent\",
  \"input\": {\"messages\": [{\"role\": \"human\", \"content\": \"what\'s the weather in sf?\"}]},
}" && curl --request POST \
--url <DEPLOY<ENT_URL>>/threads/<THREAD_ID>/runs \
--header 'Content-Type: application/json' \
--data "{
  \"assistant_id\": \"agent\",
  \"input\": {\"messages\": [{\"role\": \"human\", \"content\": \"what\'s the weather in nyc?\"}]},
  \"multitask_strategy\": \"rollback\"
}" && curl --request GET \
--url <DEPLOYMENT_URL>/threads/<THREAD_ID>/runs/<RUN_ID>/join

View run results

We can see that the thread has data only from the second run

PythonJavascriptCURL

state = await client.threads.get_state(thread["thread_id"])

for m in convert_to_messages(state["values"]["messages"]):
    m.pretty_print()

const state = await client.threads.getState(thread["thread_id"]);

for (const m of state['values']['messages']) {
  prettyPrint(m);
}

source pretty_print.sh && curl --request GET \
--url <DEPLOYMENT_URL>/threads/<THREAD_ID>/state | \
jq -c '.values.messages[]' | while read -r element; do
    type=$(echo "$element" | jq -r '.type')
    content=$(echo "$element" | jq -r '.content | if type == "array" then tostring else . end')
    pretty_print "$type" "$content"
done

Output:

================================ Human Message =================================

what's the weather in nyc?
================================== Ai Message ==================================

[{'id': 'toolu_01JzPqefao1gxwajHQ3Yh3JD', 'input': {'query': 'weather in nyc'}, 'name': 'tavily_search_results_json', 'type': 'tool_use'}]
Tool Calls:
  tavily_search_results_json (toolu_01JzPqefao1gxwajHQ3Yh3JD)
 Call ID: toolu_01JzPqefao1gxwajHQ3Yh3JD
  Args:
    query: weather in nyc
================================= Tool Message =================================
Name: tavily_search_results_json

[{"url": "https://www.weatherapi.com/", "content": "{'location': {'name': 'New York', 'region': 'New York', 'country': 'United States of America', 'lat': 40.71, 'lon': -74.01, 'tz_id': 'America/New_York', 'localtime_epoch': 1718734479, 'localtime': '2024-06-18 14:14'}, 'current': {'last_updated_epoch': 1718733600, 'last_updated': '2024-06-18 14:00', 'temp_c': 29.4, 'temp_f': 84.9, 'is_day': 1, 'condition': {'text': 'Sunny', 'icon': '//cdn.weatherapi.com/weather/64x64/day/113.png', 'code': 1000}, 'wind_mph': 2.2, 'wind_kph': 3.6, 'wind_degree': 158, 'wind_dir': 'SSE', 'pressure_mb': 1025.0, 'pressure_in': 30.26, 'precip_mm': 0.0, 'precip_in': 0.0, 'humidity': 63, 'cloud': 0, 'feelslike_c': 31.3, 'feelslike_f': 88.3, 'windchill_c': 28.3, 'windchill_f': 82.9, 'heatindex_c': 29.6, 'heatindex_f': 85.3, 'dewpoint_c': 18.4, 'dewpoint_f': 65.2, 'vis_km': 16.0, 'vis_miles': 9.0, 'uv': 7.0, 'gust_mph': 16.5, 'gust_kph': 26.5}}"}]
================================== Ai Message ==================================

The weather API results show that the current weather in New York City is sunny with a temperature of around 85°F (29°C). The wind is light at around 2-3 mph from the south-southeast. Overall it looks like a nice sunny summer day in NYC.

Verify that the original, rolled back run was deleted

PythonJavascript

try:
    await client.runs.get(thread["thread_id"], rolled_back_run["run_id"])
except httpx.HTTPStatusError as _:
    print("Original run was correctly deleted")

try {
  await client.runs.get(thread["thread_id"], rolledBackRun["run_id"]);
} catch (e) {
  console.log("Original run was correctly deleted");
}

Output:

Original run was correctly deleted

Comments