Stack Overflow Trends Monitor is an open-source research tool for discovering current Stack Overflow discussions, tracking technical topics over time, and turning question data into AI-powered reports on developer trends and pain points.
Research a specific topic or use Explore trends for a site-wide overview, then save research as a recurring monitor to track changes over time.
The tool combines the Stack Exchange API for broad discovery with headless Playwright and Decodo rotating residential proxies for full-thread research, then uses your chosen LLM provider to analyze the collected data and generate reports.
- Key features
- How it works
- Prerequisites
- Installation
- Configuration
- Site-wide trend discovery
- Tracking trends over time
- Recurring monitoring
- Research methodology
- FAQ
- Discover what developers are discussing now. Explore current Stack Overflow trends across the platform without needing to know what to search for first.
- Research specific technical topics. Enter a keyword to uncover related discussions, recurring questions, developer pain points, and emerging signals.
- Go beyond question metadata. Selected questions are researched as full Stack Overflow threads, including answers and discussion.
- Track topics over time. Save recurring monitors to identify persistent questions, recurring themes, emerging signals, and topics that disappear from later samples.
- Keep findings verifiable. Reports link themes and insights back to the Stack Overflow questions that support them.
- Export your research. Download completed reports as Markdown or JSON for further analysis or use elsewhere.
- Choose a research mode. Enter a Stack Overflow topic or use Explore trends for site-wide discovery.
- Configure the analysis. Select a trend window, research depth, maximum question count, optional tag filters, and, if needed, a recurring schedule.
- Retrieve Stack Overflow data. The Stack Exchange API collects questions for the current and comparison periods or the selected site-wide rankings.
- Filter for relevance. The LLM separates directly relevant questions from adjacent and incidental matches.
- Research selected threads. Playwright opens promising questions through Decodo rotating residential proxies and retrieves the full thread for deeper analysis.
- Generate the report. The LLM combines quantitative signals with thread evidence to identify themes, pain points, emerging signals, and notable questions.
- Track changes over time. Save the configuration as a recurring monitor to build longitudinal history across later runs.
- Export the results. Download a completed report as Markdown or JSON.
Reports can be exported as Markdown or JSON. The example below is a shortened excerpt from a Current trends research run using Thorough research depth.
View example Markdown report
# Stack Overflow Trends Report
**Topic:** Stack Overflow trends
**Run:** 2026-08-25
## Research basis
- 31 analyzed from 31 retrieved
- Scope · site-wide · week · rankings: 4/4
- Current sample · 2026-08-18 to 2026-08-25
- Analysis selection · 31 eligible · 31 selected for analysis · thorough research depth · deep dives: 15/15 successful
- Evidence depth · selected deep dives retrieved as full Stack Overflow threads through Playwright
- Top tags · c++ (7), python (4), android (2), c (2), css (2), dataframe (2), java (2), javascript (2)
### Highest engagement
- [Removing an unused class member slows down program?](https://stackoverflow.com/questions/79996455/removing-an-unused-class-member-slows-down-program) · 19 score · 1 answer · 538 views
- [Should (double)(std::float16_t(1024) + std::float16_t(1025)) be 2048.0 or 2049.0?](https://stackoverflow.com/questions/79996927/should-doublestdfloat16-t1024-stdfloat16-t1025-be-2048-0-or-2049-0) · 12 score · 3 answers · 877 views
- [How can I find a specific consecutive sequence of rows in SQL without MATCH_RECOGNIZE?](https://stackoverflow.com/questions/79997241/how-can-i-find-a-specific-consecutive-sequence-of-rows-in-sql-without-match-reco) · 5 score · 5 answers · 231 views
## Current trend summary
This week's retrieved Stack Overflow sample is dominated by C++ questions covering low-level performance behavior, floating-point precision with C++23 extended types, and compiler divergence on language-standard edge cases. Several C++ questions are the highest-engagement posts in the sample and have been present across multiple prior observation windows.
## Across previous runs
Six observation windows span August 14–25, 2026. Three C++ questions covering struct layout and 4K aliasing, `std::float16_t` precision, and `std::pair` trivial copyability first appeared in the August 21 window and remain present in the current August 25 window, representing recurrence across two overlapping windows from distinct authors.
- **C++ low-level performance and standard-library semantics** · recurring · recurring topic · high confidence · 3 distinct questions · 3 authors · observed on 4 dates
- **C++ compiler divergence on conversion function overload resolution** · recurring · recurring topic · single-author evidence · low confidence · 2 distinct questions · 1 author · observed on 3 dates
- **Java Gatherers API type inference and documentation friction** · not-observed · historical absence · low confidence · 2 distinct questions · 1 author · observed on 4 dates
## Discussion themes
### C++ language-lawyer: compiler divergence on conversion function overload resolution
Two questions from the same author investigate cases where MSVC, GCC, Clang, and EDG disagree on which conversion function to select when templated and non-templated overloads coexist. The evidence comes from a single author, which limits its strength as a signal of broad developer recurrence.
- Evidence strength · 2 distinct questions · 1 author · low evidence
- Evidence · [Better match conversion function when a templated member function is available](https://stackoverflow.com/questions/79997133/better-match-conversion-function-when-a-templated-member-function-is-available)
- Evidence · [Explicit call to conversion function treated differently than implicit call when using explicit object parameter](https://stackoverflow.com/questions/79997361/explicit-call-to-conversion-function-treated-differently-than-implicit-call-when)
...Before installing the project, make sure you have:
- Bun 1.2.5 or newer
- Docker or another Docker-compatible runtime
- Google Chrome, used by Playwright for full-thread research
- Decodo residential proxy credentials
- At least one supported LLM provider API key:
- Anthropic
- OpenAI
- Google Gemini
A Stack Exchange API key is not required for the current setup.
In your terminal, run:
git clone https://github.com/Decodo/stackoverflow-trends-monitor.git
cd stackoverflow-trends-monitorbun installThe full-thread research stage requires Decodo residential proxy credentials. To set them up:
- Register or log in to the Decodo dashboard.
- Go to Residential proxies and choose a plan or start a 3-day free trial.
- Open Proxy setup.
- Select a location or choose Random.
- Select Rotating as the session type and choose HTTP(S) as the protocol.
- Choose your authentication method.
- Copy the generated proxy username and password. You'll add these credentials to the
.envfile in the next step.
Run this command in your terminal to create a local .env file from the provided example:
cp .env.example .envFiles beginning with a dot may be hidden by default. If you don't see .env in the project folder, enable hidden files in your file browser:
- On macOS Finder, press Cmd+Shift+.
- On most Linux file managers, press Ctrl+H
- On Windows,
.envshould normally be visible in File Explorer; if it isn't, select View → Show → Hidden items
Open the newly created .env file in a text editor. Add your Decodo residential proxy username and password, then select an LLM provider and add its API key:
# Backend
PORT=5002
PUBLIC_API_BASE_URL=http://localhost:5002
PUBLIC_FRONTEND_URL=http://localhost:5274
# Database
MONGO_PORT=27018
REDIS_PORT=6378
MONGODB_URI=mongodb://localhost:27018/platform
# Decodo residential proxy
DECODO_PROXY_USERNAME=
DECODO_PROXY_PASSWORD=
DECODO_PROXY_HOST=gate.decodo.com
DECODO_PROXY_PORT=7000
# Playwright
PLAYWRIGHT_CHANNEL=chrome
# LLM provider: claude | openai | gemini
LLM_PROVIDER=claude
LLM_MODEL=
# API keys
ANTHROPIC_API_KEY=
OPENAI_API_KEY=
GEMINI_API_KEY=Only the API key for your selected LLM_PROVIDER is required. For example, if you use LLM_PROVIDER=claude, add your ANTHROPIC_API_KEY and leave the OpenAI and Gemini keys empty.
Make sure Docker Desktop (or another Docker-compatible runtime) is running, then start MongoDB and Redis with this command in your terminal:
bun db:upbun run buildbun devKeep this terminal window open while using the application. Closing it stops the local frontend and backend.
To reach the frontend, open your browser and go to:
http://localhost:5274
The backend API runs at:
http://localhost:5002
| Setting | Best for |
|---|---|
| Focused | Faster research with fewer full-thread deep dives |
| Standard | Balanced everyday analysis and the default setting |
| Thorough | Broader secondary evidence |
| Comprehensive | Maximum useful coverage for deeper research |
Higher research depth can increase analysis time, proxy usage, and LLM usage.
Tag filtering is disabled by default. Enable it when a broad topic needs to be constrained to one or more Stack Overflow tags.
For example, researching python with the tags python and django
returns a much narrower sample than researching the keyword alone. Tags
are explicit user filters and are not automatically selected by the LLM.
Explore trends requires no keyword. It combines four Stack Exchange question rankings:
| Ranking | Purpose |
|---|---|
| Hot | Questions currently ranked by Stack Overflow's Hot formula |
| Week | Questions in the weekly ranking |
| Votes | Highest-voted questions created during the current week |
| Activity | Recently active questions created during the current week |
The tool also identifies frequently occurring tags in the weekly sample, deep-dives into selected threads, and generates a broader report from the retrieved trend sample.
The resulting metrics describe the retrieved sample rather than total Stack Overflow activity.
Topic research compares two equivalent periods. A Past week analysis compares the current seven-day window with the preceding seven days, while Past month uses equivalent monthly windows.
The report separates retrieved-sample activity from broader Stack Overflow traffic. If collection reaches the configured cap, counts are shown as lower bounds and percentage-change claims are suppressed rather than extrapolated from incomplete data.
Recurring monitors add a second layer of analysis. Stored reports from earlier observation dates let the tool distinguish between:
- Persistent questions, where the same Stack Overflow post remains visible across multiple runs.
- Recurring topics, where independent questions about the same issue appear across different observation dates.
- Emerging signals, where a coherent issue appears in the current sample but does not yet have enough history to establish recurrence.
- Historical absence, where a previously observed topic is not present in the current sample.
Repeated runs on the same calendar day are collapsed into one longitudinal observation window so testing does not artificially strengthen a trend.
Any topic analysis can be saved as a recurring monitor. A monitor stores its research configuration, including the topic, trend window, research depth, question limit, tag filters, and cadence.
Supported schedules include:
- Custom intervals in hours
- Daily
- Weekly
- Monthly
The local NestJS backend checks for due monitors once per minute. Automatic runs occur only while the backend and MongoDB are running, so the included scheduler is intended for local or always-on deployments rather than offline execution.
The Monitors view shows the saved configuration, next and previous run times, and the status of the latest run. Monitors can also be run manually, paused, or deleted.
Broad Stack Overflow searches can contain incidental keyword matches, so retrieved questions are classified before the report is synthesized. Incidental matches are excluded from the main analysis, and low-precision searches are flagged when fewer than half of retrieved questions are directly relevant.
Established discussion themes and developer pain points require evidence from multiple directly relevant questions. A single unusual question can still appear as a notable question, but it is not promoted into a recurring theme without supporting evidence.
For deeper research, the tool ranks useful questions by research value and topical diversity. Playwright then opens a research-depth-dependent selection of full Stack Overflow threads. A deep dive succeeds only after the actual question DOM is available and any Cloudflare challenge has cleared. Failed pages are retried with fresh browser contexts.
Reports disclose how many deep dives were attempted and completed. If no full-thread retrieval succeeds, the report explicitly falls back to API-only evidence.
| Decodo Stack Overflow Trends Monitor | Official Stack Overflow Trends | Stack Exchange Data Explorer | |
|---|---|---|---|
| Current status | Open-source and self-hosted | Retired in March 2026 | Available |
| Topic monitoring | Yes | Historical tag trends | Requires custom queries |
| Recurring monitoring | Yes | No | No built-in monitor workflow |
| AI analysis | Yes | No | No |
| Full-thread research | Yes | No | Query-dependent |
| Longitudinal reports | Yes | No | Requires custom analysis |
| Markdown and JSON export | Yes | No | Query results |
Stack Overflow Trends Monitor is not a replacement for every Data Explorer use case. Data Explorer remains better suited to custom SQL-style analysis of Stack Exchange data, while this project focuses on repeatable topic research, monitoring, and evidence-based report generation.
The tool uses a hybrid retrieval strategy.
The Stack Exchange API handles broad question discovery, comparison periods, and site-wide rankings. The client honors API-requested backoff periods, caches identical requests briefly, and retries transient throttling or availability errors.
For deeper evidence, Playwright opens selected Stack Overflow question pages through Decodo rotating residential proxies. Stack Overflow pages may initially present a Cloudflare challenge. The scraper waits for the real question DOM and retries failed pages with fresh browser contexts rather than treating the initial response as usable thread content.
This approach keeps broad discovery efficient while allowing the report generator to analyze selected answers, comments, and full question context.
| Layer | Technology |
|---|---|
| Frontend | React 19, TanStack Router, TanStack Query, Tailwind CSS v4, Radix UI |
| Backend | NestJS 11, MongoDB, Mongoose |
| Discovery | Stack Exchange API |
| Deep-dive retrieval | Playwright, Decodo rotating residential proxies |
| LLMs | Anthropic Claude, OpenAI GPT, Google Gemini |
| Runtime | Bun |
| Local services | Docker Compose, MongoDB, Redis |
apps/
backend/src/features/
stackexchange/ Stack Exchange API discovery and rankings
decodo/ Playwright thread retrieval through Decodo proxies
tracker/ Topic comparison, trend discovery, deep dives, and reporting
monitors/ Saved schedules and recurring execution
queries/ Run history and longitudinal retrieval
llm/ LLM provider abstraction and prompts
settings/ API and model configuration
frontend/src/features/
tracker/ Topic analysis, site-wide trends, and reports
monitors/ Recurring monitor controls
queries/ History
settings/ Runtime configuration
| Command | Description |
|---|---|
bun dev |
Start frontend and backend development servers |
bun run build |
Build all application packages |
bun lint |
Run linting across all packages |
bun db:up |
Start MongoDB and Redis through Docker Compose |
bun db:down |
Stop local database containers |
No. Stack Overflow retired its official Trends tool in March 2026. The Decodo Stack Overflow Trends Monitor is an open-source alternative focused on current topic research, recurring monitoring, question-volume comparisons, AI-generated analysis, and developer pain point detection rather than reproducing the retired interface exactly.
Enter a broad topic, optionally constrain it with Stack Overflow tags, and choose a weekly or monthly comparison window. Stack Overflow Trends Monitor retrieves equivalent current and previous samples, reports activity and tag signals, and can save the configuration as a recurring monitor to build longitudinal history across later runs.
Both. The Stack Exchange API handles broad question discovery, period comparisons, and site-wide rankings. Selected questions are then opened with a custom Playwright scraper through Decodo rotating residential proxies so the analysis can use full thread context, including answers and discussion that are not represented by basic question metadata alone.
The Stack Exchange API can request a backoff period or otherwise throttle requests. The client honors requested backoff periods, caches identical requests for one minute, and retries transient throttling or availability failures. Site-wide discovery can continue when an individual ranking source is temporarily unavailable, and the report shows how many ranking sources were retrieved.
Yes. The Stack Exchange API provides question discovery and metadata, but the tool needs access to full Stack Overflow threads to analyze answers, comments, and discussion in depth. Playwright retrieves these pages through Decodo rotating residential proxies, which provide reliable access when direct automated requests encounter anti-bot protections. Without the residential proxy connection, the tool is limited to API data and cannot perform its full-thread deep dives.
The Developer Survey is a periodic survey of developers and their tools, preferences, and demographics. Stack Overflow Trends Monitor instead researches questions appearing on Stack Overflow during user-selected time windows, making it useful for monitoring current technical discussions and recurring developer problems rather than survey responses.
MIT – see LICENSE.