Hass/ha-text-ai
1
0
Fork 0
mirror of https://github.com/smkrv/ha-text-ai.git synced 2026-07-29 19:23:57 +08:00
Code Issues Packages Projects Releases Wiki Activity

Release v2.0.0

Browse Source
This commit is contained in:
SMKRV
2024-11-20 01:02:27 +03:00
parent 962a089bf4
commit 5037622fb6
10 changed files with 967 additions and 324 deletions
Download Patch File Download Diff File Expand all files Collapse all files
custom_components/ha_text_ai/services.yaml
+45 -15
View File
@@ -4,6 +4,7 @@ ask_question:
Send a question to the AI model and receive a detailed response.
The response will be stored in the conversation history and can be retrieved later.
Response time may vary based on model selection and server load.
Supports various AI models with different capabilities and pricing.
fields:
question:
name: Question
@@ -11,6 +12,7 @@ ask_question:
Your question or prompt for the AI assistant. Be specific and clear for better results.
You can ask about home automation, technical advice, or general questions.
For complex queries, consider breaking them into smaller parts.
The system will maintain conversation context for follow-up questions.
required: true
example: |
What automations would you recommend for a smart kitchen?
@@ -28,8 +30,10 @@ ask_question:
name: Model
description: >-
Select an AI model to use (optional, overrides default setting).
You can choose from predefined models or enter your own model identifier.
Different models have different capabilities and token limits.
Different models have different capabilities, context limits, and response characteristics:
- GPT-3.5 Turbo: Fast, efficient, good for most tasks
- GPT-4: More capable, better reasoning, slower
- Claude-3: Advanced capabilities, longer context
Note: More capable models may have longer response times and higher API costs.
required: false
example: "gpt-3.5-turbo"
@@ -48,14 +52,16 @@ ask_question:
value: "gpt-4-32k"
- label: "GPT-4 Turbo (Latest)"
value: "gpt-4-1106-preview"
- label: "Claude-3 Sonnet"
- label: "Claude-3 Sonnet (Balanced)"
value: "claude-3-sonnet"
- label: "Claude-3 Opus (Most Advanced)"
value: "claude-3-opus"
mode: dropdown
temperature:
name: Temperature
description: >-
Controls response creativity (0-2):
Controls response creativity (0.0-2.0):
0.0-0.3: Focused, consistent responses (best for technical/factual queries)
0.4-0.7: Balanced responses (recommended for most uses)
0.8-2.0: More creative, varied responses (best for brainstorming)
@@ -73,12 +79,12 @@ ask_question:
max_tokens:
name: Max Tokens
description: >-
Maximum length of the response. Higher values allow longer responses but use more API tokens.
Maximum length of the response (1-4096 tokens).
Recommended ranges:
- Short responses (256-512): Quick answers, status updates
- Medium responses (512-1024): Detailed explanations, instructions
- Long responses (1024-4096): Complex analysis, multiple examples
Note: Actual response length may be shorter based on content.
- Short (256-512): Quick answers, status updates
- Medium (512-1024): Detailed explanations
- Long (1024-4096): Complex analysis
Note: Token limits vary by model. Actual length may be shorter.
required: false
default: 1000
selector:
@@ -93,8 +99,9 @@ ask_question:
description: >-
Optional system prompt to set context for this specific question.
This will temporarily override the default system prompt.
Use this to specify expertise areas, response format, or special instructions.
required: false
example: "You are a home automation expert focused on energy efficiency"
example: "You are a home automation expert focused on energy efficiency and security"
selector:
text:
multiline: true
@@ -104,14 +111,18 @@ clear_history:
description: >-
Delete all stored questions and responses from the conversation history.
This action cannot be undone. Consider using 'get_history' first if you need to backup the data.
System prompt settings will be preserved.
System prompt settings and configuration will be preserved.
fields: {}
get_history:
name: Get History
description: >-
Retrieve recent conversation history, including questions, responses, and timestamps.
Results are ordered from newest to oldest and include metadata like model used and response times.
Retrieve recent conversation history, including:
- Questions and responses
- Timestamps and response times
- Models used and token counts
- Temperature and other settings
Results are ordered from newest to oldest.
fields:
limit:
name: Limit
@@ -144,22 +155,41 @@ get_history:
value: "gpt-3.5-turbo"
- label: "GPT-4"
value: "gpt-4"
- label: "GPT-4 Turbo"
value: "gpt-4-1106-preview"
- label: "Claude-3 Sonnet"
value: "claude-3-sonnet"
- label: "Claude-3 Opus"
value: "claude-3-opus"
mode: dropdown
start_date:
name: Start Date
description: >-
Optional start date for filtering history.
Format: YYYY-MM-DD
required: false
selector:
datetime:
type: date
set_system_prompt:
name: Set System Prompt
description: >-
Set default system behavior instructions for all future conversations.
This defines how the AI should behave and respond to questions.
The prompt will be used for all models unless overridden per question.
fields:
prompt:
name: System Prompt
description: >-
Instructions that define how the AI should behave and respond.
Be specific about the desired expertise, tone, and format of responses.
Maximum length: 1000 characters.
Be specific about:
- Desired expertise and knowledge areas
- Response tone and style
- Output format preferences
- Special handling instructions
Maximum length: 1000 characters
required: true
example: |
You are a home automation expert assistant. Focus on: