Half-finished org-mode conversions (at least with streaming on) #813
Labels
bug
Something isn't working
Comments
I can compare the two and try to fix the conversion bug. |
|
[
{
"candidates": [
{
"content": {
"parts": [
{
"text": "Okay"
}
],
"role": "model"
}
}
],
"usageMetadata": {
"promptTokenCount": 16363,
"totalTokenCount": 16363,
"promptTokensDetails": [
{
"modality": "TEXT",
"tokenCount": 16363
}
]
},
"modelVersion": "gemini-2.0-flash"
},
{
"candidates": [
{
"content": {
"parts": [
{
"text": ", that's a fascinating and relevant topic, especially given the increasing accessibility of powerful personal"
}
],
"role": "model"
},
"safetyRatings": [
{
"category": "HARM_CATEGORY_HATE_SPEECH",
"probability": "NEGLIGIBLE"
},
{
"category": "HARM_CATEGORY_DANGEROUS_CONTENT",
"probability": "NEGLIGIBLE"
},
{
"category": "HARM_CATEGORY_HARASSMENT",
"probability": "NEGLIGIBLE"
},
{
"category": "HARM_CATEGORY_SEXUALLY_EXPLICIT",
"probability": "NEGLIGIBLE"
}
]
}
],
"usageMetadata": {
"promptTokenCount": 16363,
"totalTokenCount": 16363,
"promptTokensDetails": [
{
"modality": "TEXT",
"tokenCount": 16363
}
]
},
"modelVersion": "gemini-2.0-flash"
},
{
"candidates": [
{
"content": {
"parts": [
{
"text": " computing hardware. To research the performance of local large language models (LLMs) on"
}
],
"role": "model"
},
"safetyRatings": [
{
"category": "HARM_CATEGORY_HATE_SPEECH",
"probability": "NEGLIGIBLE"
},
{
"category": "HARM_CATEGORY_DANGEROUS_CONTENT",
"probability": "NEGLIGIBLE"
},
{
"category": "HARM_CATEGORY_HARASSMENT",
"probability": "NEGLIGIBLE"
},
{
"category": "HARM_CATEGORY_SEXUALLY_EXPLICIT",
"probability": "NEGLIGIBLE"
}
]
}
],
"usageMetadata": {
"promptTokenCount": 16363,
"totalTokenCount": 16363,
"promptTokensDetails": [
{
"modality": "TEXT",
"tokenCount": 16363
}
]
},
"modelVersion": "gemini-2.0-flash"
},
{
"candidates": [
{
"content": {
"parts": [
{
"text": " the M1 Mac Studio Ultra, I propose the following plan:\n\n1. *Gather Search Terms:* Compile a list of relevant search terms to use with"
}
],
"role": "model"
},
"safetyRatings": [
{
"category": "HARM_CATEGORY_HATE_SPEECH",
"probability": "NEGLIGIBLE"
},
{
"category": "HARM_CATEGORY_DANGEROUS_CONTENT",
"probability": "NEGLIGIBLE"
},
{
"category": "HARM_CATEGORY_HARASSMENT",
"probability": "NEGLIGIBLE"
},
{
"category": "HARM_CATEGORY_SEXUALLY_EXPLICIT",
"probability": "NEGLIGIBLE"
}
]
}
],
"usageMetadata": {
"promptTokenCount": 16363,
"totalTokenCount": 16363,
"promptTokensDetails": [
{
"modality": "TEXT",
"tokenCount": 16363
}
]
},
"modelVersion": "gemini-2.0-flash"
},
{
"candidates": [
{
"content": {
"parts": [
{
"text": " the =internet_search= tool. These terms should include specific LLMs (e.g., Llama, Alpaca, GPT-J), the M1"
}
],
"role": "model"
},
"safetyRatings": [
{
"category": "HARM_CATEGORY_HATE_SPEECH",
"probability": "NEGLIGIBLE"
},
{
"category": "HARM_CATEGORY_DANGEROUS_CONTENT",
"probability": "NEGLIGIBLE"
},
{
"category": "HARM_CATEGORY_HARASSMENT",
"probability": "NEGLIGIBLE"
},
{
"category": "HARM_CATEGORY_SEXUALLY_EXPLICIT",
"probability": "NEGLIGIBLE"
}
]
}
],
"usageMetadata": {
"promptTokenCount": 16363,
"totalTokenCount": 16363,
"promptTokensDetails": [
{
"modality": "TEXT",
"tokenCount": 16363
}
]
},
"modelVersion": "gemini-2.0-flash"
},
{
"candidates": [
{
"content": {
"parts": [
{
"text": " Mac Studio Ultra, and performance metrics (e.g., inference speed, memory usage, quantization).\n2. *Initial Search:* Perform initial searches using a broad range of search terms to identify relevant articles, blog posts, forum discussions"
}
],
"role": "model"
},
"safetyRatings": [
{
"category": "HARM_CATEGORY_HATE_SPEECH",
"probability": "NEGLIGIBLE"
},
{
"category": "HARM_CATEGORY_DANGEROUS_CONTENT",
"probability": "NEGLIGIBLE"
},
{
"category": "HARM_CATEGORY_HARASSMENT",
"probability": "NEGLIGIBLE"
},
{
"category": "HARM_CATEGORY_SEXUALLY_EXPLICIT",
"probability": "NEGLIGIBLE"
}
]
}
],
"usageMetadata": {
"promptTokenCount": 16363,
"totalTokenCount": 16363,
"promptTokensDetails": [
{
"modality": "TEXT",
"tokenCount": 16363
}
]
},
"modelVersion": "gemini-2.0-flash"
},
{
"candidates": [
{
"content": {
"parts": [
{
"text": ", and GitHub repositories.\n3. *Filter and Prioritize Results:* Analyze the search results to identify the most credible and informative sources. Prioritize results that include quantitative performance data (e.g., benchmarks, latency measurements).\n4"
}
],
"role": "model"
},
"safetyRatings": [
{
"category": "HARM_CATEGORY_HATE_SPEECH",
"probability": "NEGLIGIBLE"
},
{
"category": "HARM_CATEGORY_DANGEROUS_CONTENT",
"probability": "NEGLIGIBLE"
},
{
"category": "HARM_CATEGORY_HARASSMENT",
"probability": "NEGLIGIBLE"
},
{
"category": "HARM_CATEGORY_SEXUALLY_EXPLICIT",
"probability": "NEGLIGIBLE"
}
]
}
],
"usageMetadata": {
"promptTokenCount": 16363,
"totalTokenCount": 16363,
"promptTokensDetails": [
{
"modality": "TEXT",
"tokenCount": 16363
}
]
},
"modelVersion": "gemini-2.0-flash"
},
{
"candidates": [
{
"content": {
"parts": [
{
"text": ". *Deep Dive into Key Resources:* Use the =read_website= tool to extract and analyze the content of the most promising articles and blog posts. Summarize the key findings and performance metrics.\n5. *GitHub Exploration:* If the search results reveal relevant GitHub repositories, examine the code and documentation to understand how LL"
}
],
"role": "model"
},
"safetyRatings": [
{
"category": "HARM_CATEGORY_HATE_SPEECH",
"probability": "NEGLIGIBLE"
},
{
"category": "HARM_CATEGORY_DANGEROUS_CONTENT",
"probability": "NEGLIGIBLE"
},
{
"category": "HARM_CATEGORY_HARASSMENT",
"probability": "NEGLIGIBLE"
},
{
"category": "HARM_CATEGORY_SEXUALLY_EXPLICIT",
"probability": "NEGLIGIBLE"
}
]
}
],
"usageMetadata": {
"promptTokenCount": 16363,
"totalTokenCount": 16363,
"promptTokensDetails": [
{
"modality": "TEXT",
"tokenCount": 16363
}
]
},
"modelVersion": "gemini-2.0-flash"
},
{
"candidates": [
{
"content": {
"parts": [
{
"text": "Ms are being optimized for the M1 Mac Studio Ultra.\n6. *Synthesize Findings:* Combine the information gathered from different sources to create a comprehensive overview of the performance of local LLMs on the M1 Mac Studio Ultra. Identify any trends, limitations, or areas for further research.\n\nDoes this plan"
}
],
"
7E64
role": "model"
},
"safetyRatings": [
{
"category": "HARM_CATEGORY_HATE_SPEECH",
"probability": "NEGLIGIBLE"
},
{
"category": "HARM_CATEGORY_DANGEROUS_CONTENT",
"probability": "NEGLIGIBLE"
},
{
"category": "HARM_CATEGORY_HARASSMENT",
"probability": "NEGLIGIBLE"
},
{
"category": "HARM_CATEGORY_SEXUALLY_EXPLICIT",
"probability": "NEGLIGIBLE"
}
]
}
],
"usageMetadata": {
"promptTokenCount": 16363,
"totalTokenCount": 16363,
"promptTokensDetails": [
{
"modality": "TEXT",
"tokenCount": 16363
}
]
},
"modelVersion": "gemini-2.0-flash"
},
{
"candidates": [
{
"content": {
"parts": [
{
"text": " sound good to you? If so, I'll start by generating some search terms.\n"
}
],
"role": "model"
},
"finishReason": "STOP",
"safetyRatings": [
{
"category": "HARM_CATEGORY_HATE_SPEECH",
"probability": "NEGLIGIBLE"
},
{
"category": "HARM_CATEGORY_DANGEROUS_CONTENT",
"probability": "NEGLIGIBLE"
},
{
"category": "HARM_CATEGORY_HARASSMENT",
"probability": "NEGLIGIBLE"
},
{
"category": "HARM_CATEGORY_SEXUALLY_EXPLICIT",
"probability": "NEGLIGIBLE"
}
]
}
],
"usageMetadata": {
"promptTokenCount": 15518,
"candidatesTokenCount": 347,
"totalTokenCount": 15865,
"promptTokensDetails": [
{
"modality": "TEXT",
"tokenCount": 15518
}
],
"candidatesTokensDetails": [
{
"modality": "TEXT",
"tokenCount": 347
}
]
},
"modelVersion": "gemini-2.0-flash"
}
]And here's the corresponding output: |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Please update gptel first -- errors are often fixed by the time they're reported.
Bug Description
With org-mode enabled in gptel buffers, sometimes streamed output from the LLM is partially converted --- for instance, a common problem is half-converted italics notation, so I end up with something like "/foo*" all over the place. There aren't any errors, as far as I can tell, just this issue.
Backend
None
Steps to Reproduce
Not sure how to repro this specifically, since it happens all the time for me.
Additional Context
Emacs version: GNU Emacs 31.0.50 (build 2, x86_64-pc-linux-gnu, GTK+ Version 3.24.49, cairo version 1.18.2) of 2025-04-20
Backtrace
Log Information
The text was updated successfully, but these errors were encountered: