quge007/ai-infrastructure-problem-observatory

AI Infrastructure Problem Observatory: Topics Dataset

Quick Summary

87,718 AI infrastructure topics from GitHub issues/discussions, technical forums, and industry blog posts. Covers 23 monthly partitions (2024-12 to 2026-10). Collected original records are embedded—no external joins needed.

A topic is not automatically a confirmed problem: records may also be questions, announcements, tutorials, or other content. This release includes only the 19 problem-observation sources (12 repositories, three forums, and four blogs). The four MAP study versions are context material and are excluded from the topic corpus.

Use this dataset to:

• Understand production challenges faced by AI teams (infrastructure, deployment, optimization)
• Identify emerging problems and their temporal trends
• Build problem-aware systems (chatbots, documentation, product features)
• Research AI infrastructure needs and solution adoption patterns

September now contains 4,293 topics; October adds 5,332 topics. These are complete calendar-month collection runs within the registered scope, subject to the collection gaps below.

Dataset Overview

Metric	Value
Total Topics	87,718
Time Period	2024-12 to 2026-10 (23 months)
Size	550,678,602 Parquet bytes (550.68 MB decimal; zstd compressed)
Sources	19 observation sources across 3 types (GitHub, forums, blogs)
Data Preservation	Complete collected source records embedded; not a claim of gap-free collection

Data Breakdown

Source	Topics	Examples
GitHub	78,360	Issues/discussions from vllm, dify, litellm, sglang, ragflow, ollama, langfuse, openhands, tensorrt, lmcache, langgraph, dynamo
Technical Forums	7,913	HuggingFace Discussions, Ray Discourse, Kubernetes Discuss
Industry Blogs	1,445	Official posts from Anthropic Engineering (24), Anthropic Research (31), Manus (124), Together (158)

What's In This Dataset?

Each row represents one topic: one GitHub issue/discussion or forum thread, or one consolidated topic extracted from a blog article.

Core Content

• topic_text — Original issue/discussion/forum body, or blog-topic text reconstructed from source passages; may be empty (345 records)
• topic_title — Original thread title, or a generated complete-sentence title for a blog topic; present for every topic
• source_id — Originating source (e.g., "github_vllm", "forum_ray", "blog_anthropic")
• source_record_id — Original record ID (issue #, post ID, URL)
• source_date — Source timestamp stored as timestamp[us, tz=UTC], used for monthly partitioning

Full Source Preservation

• raw_source_record_json — Complete original record (JSON-encoded):
  • For GitHub/forums: Complete collected record, including captured comments/replies and metadata
  • For blogs: Complete article with raw HTML body
  • Use this field for original context; preservation of collected fields does not guarantee every source interaction was observable

Metadata

• source_url — Direct link to original source

This is the uncoded topics release. topic_type and topic_business_category are null placeholders. Cluster (_0) and confidence columns are not included.

Full schema: See schemas/topics_schema.json (52 columns total)

Why 100% Original Data?

This dataset includes complete source records because:

• Extracted topics may lose context; the original source is the source of truth
• Researchers often need the full conversation, not just the problem statement
• Complete data enables reproducibility and custom re-analysis
• Supports fact-checking and context verification

Result: Each topic is independently publishable — no need to reference original sources separately.

How to Use

The examples below use downloaded repository files and require pyarrow and pandas:

pip install pyarrow pandas

Load a Single Month

import pyarrow.parquet as pq

# Load September 2026 data
table = pq.read_table("database/topics/2026-09/part-00000.parquet")
print(f"Records in 2026-09: {table.num_rows}")

# Convert to pandas for exploration
df = table.to_pandas()
print(df[["topic_title", "source_id", "source_date"]].head(10))

Access Complete Original Data

import json

# Get the full original record for any topic
row = df.iloc[0]
original_source = json.loads(row["raw_source_record_json"])

# For GitHub issues: access all original fields
if row["source_id"].startswith("github"):
    print(f"Issue title: {original_source.get('title')}")
    print(f"Author: {original_source.get('author')}")
    print(f"Captured comments: {original_source.get('comments')}")

# For blog articles: raw HTML body
if row["source_id"].startswith("blog"):
    html_body = original_source.get("body") or ""
    print(f"Article length: {len(html_body)} chars")

Filter by Source Type

# GitHub issues/discussions only
github_topics = df[df["source_id"].str.contains("github", case=False)]
print(f"GitHub topics: {len(github_topics)}")

# Forums only
forum_topics = df[df["source_id"].str.startswith("forum_")]
print(f"Forum topics: {len(forum_topics)}")

# Blog articles only
blog_topics = df[df["source_id"].str.contains("blog_", case=False)]
print(f"Blog topics: {len(blog_topics)}")

Temporal Analysis

# Count topics by month
monthly_counts = df.groupby(df["source_date"].dt.strftime("%Y-%m")).size()
print(monthly_counts)

# Find topics reported in specific time ranges
import pandas as pd

recent = df[df["source_date"] >= pd.Timestamp("2026-06-01", tz="UTC")]
print(f"Topics from June 2026 onwards: {len(recent)}")

Data Quality & Completeness

✅ Verified preservation of collected records:

• GitHub/Forum: All records include original API JSON (100% fields preserved, 0 null values in raw_source_record_json)
• Blog articles: 337 articles successfully extracted and consolidated with raw HTML body
• Observation-source scope: All 87,718 retained topics are included without problem-type or confidence filtering; the four context-only MAP records are excluded

⚠️ Known characteristics:

• Blog extraction: Uses LLM-based extraction; potential for semantic interpretation variance. This release has no confidence indicators
• Multi-topic articles: Blog articles may yield multiple topics; this is intentional (preserves distinct problems in single article)
• Temporal coverage: Frozen collection window, represented by UTC month partitions (2024-12 to 2026-10); no real-time updates
• Collection gaps: The October collection retains two inaccessible blog URLs and 32 Together pages without reliable publication dates; unresolved pages are excluded from dated topics. Historical retrieval/date gaps also remain. Original-data preservation is not proof of complete source coverage

Data Collection Methodology

Collection Window: The historical collection starts at 2025-01-01 00:00:00 (Asia/Shanghai). September and October 2026 were collected through their complete UTC calendar months, ending at 2026-10-31 23:59:59.999999 UTC. Source timestamps and partition names use UTC. The 12 records in 2024-12 fall on the January 1 local-time boundary; the first UTC month is partial.

GitHub Issues & Discussions

• Automated API collection from 12 key AI infrastructure repositories
• All public issues and discussions during window
• Repositories: vllm, dify, litellm, sglang, ragflow, ollama, langfuse, openhands, tensorrt, lmcache, langgraph, dynamo

Technical Forums

• HuggingFace Discussions, Ray Discourse, Kubernetes Discuss
• All public forum topics during window

Industry Blog Posts

• 337 official technical blog articles collected and processed
• Sources: Anthropic Engineering, Anthropic Research, Manus, Together
• Covers product updates, architecture decisions, lessons learned

Topic Extraction

• GitHub/Forums: One issue, discussion, or forum thread → one topic; replies remain conversation evidence rather than additional topic rows
• Blog articles: LLM-assisted extraction of distinct topics, consolidated by title within each article
• Result: 87,718 topics with collected source records embedded

Limitations & Caveats

1. Language: Majority English-language; other languages underrepresented
2. Deduplication: Same problem discussed in multiple sources appears as separate rows (preserves frequency signal; not semantically deduplicated)
3. Temporal: This release ends at 2026-10-31 UTC; no real-time updates
4. Team attribution: No inferred team/company labels are provided. Repository names and author accounts must not automatically be treated as team identities
5. Blog content: LLM extraction may introduce minor semantic variation vs. original text

Citation

If you use this dataset, please cite:

@dataset{ltp2026_ipo_topics,
  title = {AI Infrastructure Problem Observatory: Topics Dataset},
  author = {{LTP Research Team}},
  year = {2026},
  url = {https://huggingface.co/datasets/quge007/ai-infrastructure-problem-observatory},
  license = {CC-BY-4.0}
}

Or use the provided CITATION.cff file for other formats.

License

This dataset is released under Creative Commons Attribution 4.0 International (CC-BY-4.0).

You are free to:

• Share and use this dataset for any purpose (commercial, research, education)
• Create derived works and adaptations
• Combine with other datasets

You must:

• Give appropriate credit to the dataset creators
• Link to the license
• Indicate if modifications were made

See LICENSE file for full legal text.

Getting Help

Common Questions

Q: How do I use this with my AI project? A: Load one month at a time using PyArrow, filter by source type, and extract the raw_source_record_json field for original context. See "How to Use" section above.

Q: Can I redistribute this dataset? A: Yes, under CC-BY-4.0. You must attribute the original creators and link to the license.

Q: Why are some problems duplicated across rows? A: Intentional. If one problem is discussed in both GitHub and a blog post, it appears twice. This preserves frequency information and multiple perspectives on the same issue.

Q: How do I understand the data fields? A: See schemas/topics_schema.json for complete field definitions and types.

Q: Where do I find recent topics? A: Most recent data is in database/topics/2026-10/part-00000.parquet (October 2026). Load and explore by source_date.

---

Last Updated: 2026-11-06

Data Version: 1.1.0

Schema Version: topics.v2