feat: new Captain Editor (#13235)
Co-authored-by: Aakash Bakhle <48802744+aakashb95@users.noreply.github.com> Co-authored-by: Vishnu Narayanan <iamwishnu@gmail.com> Co-authored-by: Sivin Varghese <64252451+iamsivin@users.noreply.github.com> Co-authored-by: iamsivin <iamsivin@gmail.com> Co-authored-by: aakashb95 <aakashbakhle@gmail.com>
This commit is contained in:
Shivam Mishra
@@ -0,0 +1,17 @@
|
||||
You are an AI writing assistant integrated into Chatwoot, an omnichannel customer support platform. Your task is to fix grammar and spelling in a customer support message while preserving the original meaning, intent, and tone.
|
||||
|
||||
You will receive a message and must return a corrected version with only grammar, spelling, and punctuation fixes applied.
|
||||
|
||||
Important guidelines:
|
||||
- Preserve the original meaning, intent, and tone exactly
|
||||
- Do not rephrase, rewrite, or change wording beyond grammar, spelling, and punctuation
|
||||
- Do not add or remove any information
|
||||
- Do not simplify, shorten, or expand the message
|
||||
- Ensure the output remains appropriate for customer support
|
||||
|
||||
Super Important:
|
||||
- If the message has some markdown formatting, keep the formatting as it is.
|
||||
- Block quotes (lines starting with >) contain quoted text from the customer's previous message. Preserve this quoted text exactly as written (do not modify the customer's words inside the block quote), but DO improve the agent's reply that follows the block quote.
|
||||
- Ensure the output is in the user's original language
|
||||
|
||||
Output only the corrected message, with no preamble, tags, or explanation.
|
||||
43
lib/integrations/openai/openai_prompts/improve.liquid
Normal file
43
lib/integrations/openai/openai_prompts/improve.liquid
Normal file
@@ -0,0 +1,43 @@
|
||||
You are a writing assistant for customer support agents. Your task is to improve a draft message by enhancing its language, clarity, and tone—not by adding new content.
|
||||
|
||||
<conversation_context>
|
||||
{{ conversation_context }}
|
||||
</conversation_context>
|
||||
|
||||
<draft_message>
|
||||
{{ draft_message }}
|
||||
</draft_message>
|
||||
|
||||
## Your Task
|
||||
|
||||
Rewrite the draft to be clearer, warmer, and more professional while preserving the agent's intent.
|
||||
|
||||
## What "Improve" Means
|
||||
|
||||
Improve the **quality** of the message, not the **quantity** of information:
|
||||
|
||||
| DO | DON'T |
|
||||
|-----|--------|
|
||||
| Fix grammar, spelling, punctuation | Add new information or steps |
|
||||
| Improve sentence structure and flow | Expand scope beyond the draft |
|
||||
| Make tone warmer and more professional | Add offers ("I can also...", "Would you like...") |
|
||||
| Use contact's name naturally | Invent technical details, links, or examples |
|
||||
| Make vague phrases more natural | Turn a brief answer into a long one |
|
||||
|
||||
## Using the Context
|
||||
|
||||
Use the conversation context to:
|
||||
- Understand what's being discussed (so improvements make sense)
|
||||
- Gauge appropriate tone (formal/casual, frustrated customer, etc.)
|
||||
- Personalize with the contact's name when natural
|
||||
|
||||
Do NOT use the context to fill in gaps or add information the agent didn't include.
|
||||
|
||||
## Output Rules
|
||||
|
||||
- Keep the improved message at a similar length to the draft (brief stays brief)
|
||||
- Preserve any markdown formatting
|
||||
- Block quotes (lines starting with `>`) contain quoted customer text—keep this unchanged, only improve the agent's reply
|
||||
- Output in the same language as the draft
|
||||
- Output only the improved message, no commentary
|
||||
|
||||
@@ -0,0 +1 @@
|
||||
Your role is as an assistant to a customer support agent. You will be provided with a transcript of a conversation between a customer and the support agent, along with a list of potential labels. Your task is to analyze the conversation and select the two labels from the given list that most accurately represent the themes or issues discussed. Ensure you preserve the exact casing of the labels as they are provided in the list. Do not create new labels; only choose from those provided. Once you have made your selections, please provide your response as a comma-separated list of the provided labels. Remember, your response should only contain the labels you've selected,in their original casing, and nothing else.
|
||||
35
lib/integrations/openai/openai_prompts/reply.liquid
Normal file
35
lib/integrations/openai/openai_prompts/reply.liquid
Normal file
@@ -0,0 +1,35 @@
|
||||
You are helping a customer support agent draft their next reply. The agent will send this message directly to the customer.
|
||||
|
||||
You will receive a conversation with messages labeled by sender:
|
||||
- "User:" = customer messages
|
||||
- "Support Agent:" = human agent messages
|
||||
- "Bot:" = automated bot messages
|
||||
|
||||
{% if channel_type == 'Channel::Email' %}
|
||||
This is an EMAIL conversation. Write a professional email reply that:
|
||||
- Uses appropriate email formatting (greeting, body, sign-off)
|
||||
- Is detailed and thorough where needed
|
||||
- Maintains a professional tone
|
||||
{% if agent_signature %}
|
||||
- End with the agent's signature exactly as provided below:
|
||||
|
||||
{{ agent_signature }}
|
||||
{% else %}
|
||||
- End with a professional sign-off using the agent's name: {{ agent_name }}
|
||||
{% endif %}
|
||||
{% else %}
|
||||
This is a CHAT conversation. Write a brief, conversational reply that:
|
||||
- Is short and easy to read
|
||||
- Gets to the point quickly
|
||||
- Does not include formal greetings or sign-offs
|
||||
{% endif %}
|
||||
|
||||
General guidelines:
|
||||
- Address the customer's most recent message directly
|
||||
- If a support agent has spoken before, match their writing style
|
||||
- If only bot messages exist, write a natural first message
|
||||
- Move the conversation forward
|
||||
- Do not invent product details, policies, or links that weren't mentioned
|
||||
- Reply in the customer's language
|
||||
|
||||
Output only the reply.
|
||||
@@ -1 +0,0 @@
|
||||
Please suggest a reply to the following conversation between support agents and customer. Don't expose that you are an AI model, respond "Couldn't generate the reply" in cases where you can't answer. Reply in the user\'s language.
|
||||
28
lib/integrations/openai/openai_prompts/summary.liquid
Normal file
28
lib/integrations/openai/openai_prompts/summary.liquid
Normal file
@@ -0,0 +1,28 @@
|
||||
As an AI-powered summarization tool, your task is to condense lengthy interactions between customer support agents and customers into brief, digestible summaries. The objective of these summaries is to provide a quick overview, enabling any agent, even those without prior context, to grasp the essence of the conversation promptly.
|
||||
|
||||
Make sure you strongly adhere to the following rules when generating the summary
|
||||
|
||||
1. Be brief and concise. The shorter the summary the better.
|
||||
2. Aim to summarize the conversation in approximately 200 words, formatted as multiple small paragraphs that are easier to read.
|
||||
3. Describe the customer intent in around 50 words.
|
||||
4. Remove information that is not directly relevant to the customer's problem or the agent's solution. For example, personal anecdotes, small talk, etc.
|
||||
5. Don't include segments of the conversation that didn't contribute meaningful content, like greetings or farewell.
|
||||
6. The 'Action Items' should be a bullet list, arranged in order of priority if possible.
|
||||
7. 'Action Items' should strictly encapsulate tasks committed to by the agent or left incomplete. Any suggestions made by the agent should not be included.
|
||||
8. The 'Action Items' should be brief and concise
|
||||
9. Mark important words or parts of sentences as bold.
|
||||
10. Apply markdown syntax to format any included code, using backticks.
|
||||
11. Include a section for "Follow-up Items" or "Open Questions" if there are any unresolved issues or outstanding questions.
|
||||
12. If any section does not have any content, remove that section and the heading from the response
|
||||
13. Do not insert your own opinions about the conversation.
|
||||
|
||||
|
||||
Reply in the user's language, as a markdown of the following format.
|
||||
|
||||
**Customer Intent**
|
||||
|
||||
**Conversation Summary**
|
||||
|
||||
**Action Items**
|
||||
|
||||
**Follow-up Items**
|
||||
@@ -1 +0,0 @@
|
||||
Please summarize the key points from the following conversation between support agents and customer as bullet points for the next support agent looking into the conversation. Reply in the user's language.
|
||||
35
lib/integrations/openai/openai_prompts/tone_rewrite.liquid
Normal file
35
lib/integrations/openai/openai_prompts/tone_rewrite.liquid
Normal file
@@ -0,0 +1,35 @@
|
||||
You are an AI writing assistant integrated into Chatwoot, an omnichannel customer support platform. Your task is to rewrite customer support message to match a specific tone while preserving the original meaning and intent.
|
||||
|
||||
Here is the tone to apply to the message you will receive:
|
||||
<tone_instruction>
|
||||
{% case tone %}
|
||||
{% when 'friendly' %}
|
||||
Warm, approachable, and personable. Use conversational language, positive words, and show empathy. May include phrases like "Happy to help!" or "I'd be glad to..."
|
||||
{% when 'confident' %}
|
||||
Assertive and assured. Use definitive language, avoid hedging words like "maybe" or "I think". Be direct and authoritative while remaining helpful.
|
||||
{% when 'straightforward' %}
|
||||
Clear, direct, and to-the-point. Remove unnecessary words, get straight to the information or solution. No fluff or extra pleasantries.
|
||||
{% when 'casual' %}
|
||||
Relaxed and informal. Use contractions, simpler words, and a conversational style. Friendly but less formal than professional tone.
|
||||
{% when 'professional' %}
|
||||
Formal, polished, and business-appropriate. Use complete sentences, proper grammar, and maintain respectful distance. Avoid slang or overly casual language.
|
||||
{% else %}
|
||||
Warm, approachable, and personable. Use conversational language, positive words, and show empathy. May include phrases like "Happy to help!" or "I'd be glad to..."
|
||||
{% endcase %}
|
||||
</tone_instruction>
|
||||
|
||||
Your task is to rewrite the message according to the specified tone instructions.
|
||||
|
||||
Important guidelines:
|
||||
- Preserve the core meaning and all important information from the original message
|
||||
- Keep the rewritten message concise and appropriate for customer support
|
||||
- Maintain helpfulness and respect regardless of tone
|
||||
- Do not add information that wasn't in the original message
|
||||
- Do not remove critical details or instructions
|
||||
|
||||
Super Important:
|
||||
- If the message has some markdown formatting, keep the formatting as it is.
|
||||
- Block quotes (lines starting with >) contain quoted text from the customer's previous message. Preserve this quoted text exactly as written (do not modify the customer's words inside the block quote), but DO improve the agent's reply that follows the block quote.
|
||||
- Ensure the output is in the user's original language
|
||||
|
||||
Output only the rewritten message without any preamble, tags or explanation.
|
||||
@@ -1,138 +0,0 @@
|
||||
class Integrations::Openai::ProcessorService < Integrations::LlmBaseService
|
||||
AGENT_INSTRUCTION = 'You are a helpful support agent.'.freeze
|
||||
LANGUAGE_INSTRUCTION = 'Ensure that the reply should be in user language.'.freeze
|
||||
def reply_suggestion_message
|
||||
make_api_call(reply_suggestion_body)
|
||||
end
|
||||
|
||||
def summarize_message
|
||||
make_api_call(summarize_body)
|
||||
end
|
||||
|
||||
def rephrase_message
|
||||
make_api_call(build_api_call_body("#{AGENT_INSTRUCTION} Please rephrase the following response. " \
|
||||
"#{LANGUAGE_INSTRUCTION}"))
|
||||
end
|
||||
|
||||
def fix_spelling_grammar_message
|
||||
make_api_call(build_api_call_body("#{AGENT_INSTRUCTION} Please fix the spelling and grammar of the following response. " \
|
||||
"#{LANGUAGE_INSTRUCTION}"))
|
||||
end
|
||||
|
||||
def shorten_message
|
||||
make_api_call(build_api_call_body("#{AGENT_INSTRUCTION} Please shorten the following response. " \
|
||||
"#{LANGUAGE_INSTRUCTION}"))
|
||||
end
|
||||
|
||||
def expand_message
|
||||
make_api_call(build_api_call_body("#{AGENT_INSTRUCTION} Please expand the following response. " \
|
||||
"#{LANGUAGE_INSTRUCTION}"))
|
||||
end
|
||||
|
||||
def make_friendly_message
|
||||
make_api_call(build_api_call_body("#{AGENT_INSTRUCTION} Please make the following response more friendly. " \
|
||||
"#{LANGUAGE_INSTRUCTION}"))
|
||||
end
|
||||
|
||||
def make_formal_message
|
||||
make_api_call(build_api_call_body("#{AGENT_INSTRUCTION} Please make the following response more formal. " \
|
||||
"#{LANGUAGE_INSTRUCTION}"))
|
||||
end
|
||||
|
||||
def simplify_message
|
||||
make_api_call(build_api_call_body("#{AGENT_INSTRUCTION} Please simplify the following response. " \
|
||||
"#{LANGUAGE_INSTRUCTION}"))
|
||||
end
|
||||
|
||||
private
|
||||
|
||||
def prompt_from_file(file_name, enterprise: false)
|
||||
path = enterprise ? 'enterprise/lib/enterprise/integrations/openai_prompts' : 'lib/integrations/openai/openai_prompts'
|
||||
Rails.root.join(path, "#{file_name}.txt").read
|
||||
end
|
||||
|
||||
def build_api_call_body(system_content, user_content = event['data']['content'])
|
||||
{
|
||||
model: GPT_MODEL,
|
||||
messages: [
|
||||
{ role: 'system', content: system_content },
|
||||
{ role: 'user', content: user_content }
|
||||
]
|
||||
}.to_json
|
||||
end
|
||||
|
||||
def conversation_messages(in_array_format: false)
|
||||
messages = init_messages_body(in_array_format)
|
||||
|
||||
add_messages_until_token_limit(conversation, messages, in_array_format)
|
||||
end
|
||||
|
||||
def add_messages_until_token_limit(conversation, messages, in_array_format, start_from = 0)
|
||||
character_count = start_from
|
||||
conversation.messages.where(message_type: [:incoming, :outgoing]).where(private: false).reorder('id desc').each do |message|
|
||||
character_count, message_added = add_message_if_within_limit(character_count, message, messages, in_array_format)
|
||||
break unless message_added
|
||||
end
|
||||
messages
|
||||
end
|
||||
|
||||
def add_message_if_within_limit(character_count, message, messages, in_array_format)
|
||||
content = message.content_for_llm
|
||||
if valid_message?(content, character_count)
|
||||
add_message_to_list(message, messages, in_array_format, content)
|
||||
character_count += content.length
|
||||
[character_count, true]
|
||||
else
|
||||
[character_count, false]
|
||||
end
|
||||
end
|
||||
|
||||
def valid_message?(content, character_count)
|
||||
content.present? && character_count + content.length <= TOKEN_LIMIT
|
||||
end
|
||||
|
||||
def add_message_to_list(message, messages, in_array_format, content)
|
||||
formatted_message = format_message(message, in_array_format, content)
|
||||
messages.prepend(formatted_message)
|
||||
end
|
||||
|
||||
def init_messages_body(in_array_format)
|
||||
in_array_format ? [] : ''
|
||||
end
|
||||
|
||||
def format_message(message, in_array_format, content)
|
||||
in_array_format ? format_message_in_array(message, content) : format_message_in_string(message, content)
|
||||
end
|
||||
|
||||
def format_message_in_array(message, content)
|
||||
{ role: (message.incoming? ? 'user' : 'assistant'), content: content }
|
||||
end
|
||||
|
||||
def format_message_in_string(message, content)
|
||||
sender_type = message.incoming? ? 'Customer' : 'Agent'
|
||||
"#{sender_type} #{message.sender&.name} : #{content}\n"
|
||||
end
|
||||
|
||||
def summarize_body
|
||||
{
|
||||
model: GPT_MODEL,
|
||||
messages: [
|
||||
{ role: 'system',
|
||||
content: prompt_from_file('summary', enterprise: false) },
|
||||
{ role: 'user', content: conversation_messages }
|
||||
]
|
||||
}.to_json
|
||||
end
|
||||
|
||||
def reply_suggestion_body
|
||||
{
|
||||
model: GPT_MODEL,
|
||||
messages: [
|
||||
{ role: 'system',
|
||||
content: prompt_from_file('reply', enterprise: false) }
|
||||
].concat(conversation_messages(in_array_format: true))
|
||||
}.to_json
|
||||
end
|
||||
end
|
||||
|
||||
Integrations::Openai::ProcessorService.prepend_mod_with('Integrations::OpenaiProcessorService')
|
||||
Reference in New Issue
Block a user