# AI and machine learning Kit: Apache projects for small business

October 29, 2026. [Read the article](https://jwatte.com/blog/apache-ai-machine-learning-small-business/).

## How to use this kit

Choose one output a person can check against a known answer. Keep a held-out sample of normal, incomplete and misleading inputs before trying a model. Name the developer, IT operator and person who approves consequential actions. No project in this group is classified as an owner-run application. These are documentation-based trial plans, not installed or benchmarked systems.

## Trial cards

### Texera: visual analysis with IT ownership

Fits a machine shop joining inspection and batch records, or two coffee shops reconciling sales and waste. The [Docker guide](https://texera.incubator.apache.org/docs/v1.2.0/getting-started/installing-using-docker/) specifies minimum 2 CPU cores/4 GB RAM/20 GB disk and recommends 8 cores/16 GB/50 GB. Start with the versioned Compose bundle for [1.2.0-incubating](https://texera.incubator.apache.org/releases/1.2.0-incubating/), not mixed images. Our Large VM allocation is a pilot assumption, not a CPU benchmark or high availability claim.

IT must manage access, containers, database and file backups, monitoring and upgrades. Preserve workflow definitions and results; test restoration of persistent volumes. First week: import sanitized copies, produce one known report, reconcile totals manually, restart the trial and restore from backup. Ask who can run custom code, who owns recovery and what the managed alternative's compute extras cost. Deepnote Team is the hosted notebook alternative below, not hosted Texera.

### OpenNLP: local language analysis for developers

Fits extraction from a property manager's copied maintenance notes or a retailer's stored feedback. Start with [stable 2.5.12 documentation](https://opennlp.apache.org/docs/2.5.12/manual/opennlp.html) and [official downloads](https://opennlp.apache.org/download.html). Both stable 2.5.12 and milestone 3.0.0-M6 were announced September 11, 2026; the milestone is not stable. Stable 2.x uses JDK 17; current 3.x uses JDK 21. No universal CPU/RAM floor is published; our Small CPU server requires model and batch testing.

Developers need Java, trusted models and labeled evaluation examples. [Security guidance](https://opennlp.apache.org/security.html) treats models and dictionaries as trusted assets, unlike analyzed text. Back up code, model versions and original inputs; repeat the held-out test before updates. First week: manually label a sample, run one extraction, review missed and wrong items, then decide whether it helps. Ask whether a simple rule would do and whether Comprehend's priced standard operation matches the actual task. No automatic emergency routing or compliance claim is part of this trial.

### Burr: custom workflow coordination for developers

Fits a landscaping crew's proposed follow-up tasks or a print shop's quote-form completeness checks. [Install guidance](https://burr.incubator.apache.org/docs/getting_started/install/) uses the official Python package; core Python 3.9+, optional CLI Python 3.10 or newer. No numeric hardware floor is published. Our Small machine is a coordinator allowance with no local large-model inference and no model API calls.

The [review-step documentation](https://burr.incubator.apache.org/docs/concepts/human-in-the-loop/) pauses execution; the application must obtain and validate approval. Use the [deployment guidance](https://burr.incubator.apache.org/docs/ui/deployment/) to choose protected tracking storage. Back up state, code and configuration; test retries without duplicate actions, and redact or restrict traces. First week: deterministic checks only, copied inputs, proposed outputs, and a person approves the next action. Ask who builds the permission screen and how usage is counted in a hosted alternative. Zapier pricing below is only a starting-plan floor pending a task/features quote.

## IT provider questions

Texera is the sole IT-tier entry. What fails when the one host fails? Can workflows, PostgreSQL and result files be restored consistently? Who patches images and controls user-supplied code? What measured load establishes the chosen hardware? Our shared platform allowance is 24 to 60 setup hours and 6 to 16 monthly hours, with outsourced billing requiring a quote.

## Developer reminders

- **Burr (Incubating):** a developer writes stateful Python workflows, including the review step described above; the framework does not supply a finished business assistant.
- **cTAKES:** specialized clinical-text software may use it to extract medical concepts. Before storing patient text, check HIPAA requirements and the terms for models and data; this is not a clinic chatbot or a compliance claim.
- **Magpie:** your software maintainer may use its agent-assisted triage and proposed fixes. Human review still decides what code reaches your business application.
- **Mahout:** the [current project](https://mahout.apache.org/) emphasizes Qumat and QDP for quantum machine learning. The table uses the current QuMat release instead of the stale 0.5 directory record; classic code is in maintenance mode.
- **Maka (Incubating):** a developer may build an agent workspace with action records and permissions. An activity log does not prove an action was authorized or correct.
- **OpenNLP:** your developer supplies language models and evaluates extraction errors; it is not a general answering chatbot.
- **Singa:** a model developer may use it to build and train neural networks. Training hardware and labeled data need a separate budget.
- **Sourcelume:** training-data provenance records can describe declared origin and licensing. They do not independently establish legal permission to use every item, and no numbered release was verified here.
- **SystemDS:** developers can run scripted analysis and machine-learning workflows. The owner should ask what data and test set support the resulting business prediction.
- **TVM:** a developer may compile models for particular processors. The compiler does not supply the trained business model, its training data or an accuracy guarantee.
- **UIMA:** document-processing applications can combine analysis components through this framework; Java SDK and Ruta releases have their own version numbers.

## Full group table

Registry and artifact dates are distinguished. OpenNLP's fresh release check replaces stale directory data. Incubating projects are not yet fully endorsed by the ASF; see the [incubation policy](https://incubator.apache.org/policy/incubation.html). Mahout now uses [QuMat 0.6.0, published May 31, 2026](https://github.com/apache/mahout/releases/tag/mahout-qumat-0.6.0), replacing stale directory data. Burr's publisher says August 25 while GitHub publication is August 29; the table uses the latter.

| Project | What it does | Who runs it | Latest recorded release and date |
|---|---|---|---|
| [Apache Burr (Incubating)](https://burr.incubator.apache.org/) <!-- project:incubator-burr --> | Builds observable workflows for AI agents and other state driven Python applications. | Your developer | [0.43.0-incubating; 2026-08-29](https://github.com/apache/burr/releases/tag/v0.43.0-incubating) |
| [Apache cTAKES](https://github.com/apache/ctakes) <!-- project:ctakes --> | Extracts clinical concepts from medical text for specialized software. | Your developer | [6.0.0; 2024-09-16](https://projects.apache.org/json/foundation/projects.json) (registry) |
| [Apache Magpie](https://github.com/apache/magpie) <!-- project:magpie --> | Provides an agent assisted framework for software maintainers and developers. | Your developer | [0.1.0; 2026-08-03](https://github.com/apache/magpie/releases/tag/0.1.0) |
| [Apache Mahout](https://mahout.apache.org) <!-- project:mahout --> | Provides tools for quantum machine learning and mathematical software development. | Your developer | [QuMat 0.6.0; 2026-05-31](https://github.com/apache/mahout/releases/tag/mahout-qumat-0.6.0) |
| [Apache Maka (Incubating)](https://maka.incubator.apache.org/) <!-- project:incubator-maka --> | Runs AI agent workflows and records their actions and permission decisions. | Your developer | [0.2.0-incubating; 2026-09-22](https://downloads.apache.org/incubator/maka/0.2.0-incubating/) (file date) |
| [Apache OpenNLP](https://opennlp.apache.org) <!-- project:opennlp --> | Finds language patterns, names and sentence structure in text. | Your developer | [3.0.0-M6 milestone; 2.5.12 stable; 2026-09-11](https://opennlp.apache.org/news/) |
| [Apache Singa](https://singa.apache.org) <!-- project:singa --> | Provides tools for developers building and training machine learning models. | Your developer | [5.1.0; 2026-03-25](https://downloads.apache.org/singa/5.1.0/) (file date) |
| [Apache Sourcelume](https://sourcelume.apache.org) <!-- project:sourcelume --> | Records where AI training data came from and the licensing terms its producers declare. | Your developer | Not listed |
| [Apache Systemds](https://systemds.apache.org) <!-- project:systemds --> | Runs data analysis and machine learning workflows from scripts. | Your developer | [3.4.0; 2026-07-24](https://downloads.apache.org/systemds/3.4.0/) (file date) |
| [Apache Texera (Incubating)](https://texera.incubator.apache.org/) <!-- project:incubator-texera --> | Lets teams build shared visual workflows for data analysis and machine learning. | Your IT department or provider | [1.2.0-incubating; 2026-08-14](https://github.com/apache/texera/releases/tag/v1.2.0-incubating) |
| [Apache TVM](https://tvm.apache.org/) <!-- project:tvm --> | Compiles machine learning models for different processors and accelerators. | Your developer | [0.27.0.post1; 2026-09-29](https://github.com/apache/tvm/releases/tag/v0.27.0.post1) |
| [Apache UIMA](https://uima.apache.org) <!-- project:uima --> | Combines text and other unstructured data analysis components into applications. | Your developer | [3.6.1; 2025-10-01](https://uima.apache.org/downloads.cgi) |

## Cost worksheet

Task and known answer ___. Input count and size low/high ___. Editor count low/high ___. Server allocation and node count ___. Setup hours low/high ___. Monthly admin hours low/high ___. Backup GB low/high ___. Model/API/compute extras ___. Separate development, labeling, evaluation and migration quote ___. Approval owner ___.

Prices are US dollars before tax, read October 29, 2026. Our common labor rate is $68.09/hour: the [BLS sysadmin median wage](https://data.bls.gov/oesprofile/?areas=INDUSTRY,STATE,MSA&major_group=150000&measure=01&occupation=151244&year=2025) of $47.66 divided by the [70.0% wage share](https://data.bls.gov/timeseries/CMU2020000000000P). It is an employee compensation proxy. Outside IT providers charge more, so get a quote rather than treating these figures as their tariff.

[DigitalOcean Small](https://www.digitalocean.com/pricing/droplets) is $24/month for 2 vCPU, 4 GiB RAM and 80 GiB disk. The selected Large plan is $252/month for 8 dedicated vCPU, 32 GiB RAM and 100 GiB disk. These allocations are our planning assumptions. Retained backup volume is fixed at 100 GB at both ends; [B2](https://www.backblaze.com/cloud-storage/pricing) gives (100 minus 10) x $0.00695 = $0.6255/month. No excess egress or chargeable Class D calls are assumed.

Setup = setup hours x $68.09. Monthly = host + backup + monthly administration hours x $68.09. The 36 month total = setup + 36 x monthly. Savings = alternative total minus self hosted total. Break even is ceiling(setup difference divided by monthly savings), or never when self hosting already costs more each month.

These are operation-only comparisons for prepared workflows. Bespoke development, data cleaning, annotation, model training, evaluation, integration, migration, staff training, tax and major incidents are excluded on both routes. Alternative incremental setup is zero at both ends for an existing account or prepared workflow, not a promise of free adoption. Do not use a cheap runtime calculation as a quote for building an AI product.

### Texera versus Deepnote Team

One Large server, 24 to 60 setup hours and 6 to 16 monthly hours, fixed 100 GB backups. [Deepnote Team](https://deepnote.com/pricing): fixed two editors x $39/month billed yearly, $936/year; additional compute/AI usage may cost more. Not feature-equivalent to Texera.

| Cost | Low labor case | High labor case |
|---|---:|---:|
| Setup | $1,634 | $4,085 |
| Monthly operations | $661 | $1,342 |
| Self hosted, 36 months | $25,436 | $52,400 |
| Alternative monthly | $78 | $78 |
| Alternative, 36 months | $2,808 | $2,808 |
| Savings from self hosting | Negative $22,628 | Negative $49,592 |
| Break even month | Never | Never |

### Texera setup when an AI agent does the typing

We keep Texera's Large server, 100 GB backups and two-editor comparison unchanged. Our IT planning assumptions reduce setup to 65% to 90% of manual hours and monthly administration to 80% to 100%. These are scenarios, not measured Texera deployments; coding studies do not establish server setup savings. Review already occupies 6 to 18 setup hours and 3 to 8 monthly hours within the AI totals. Prompting, testing and corrections count as human work; unattended agent time does not.

| Texera planning case | Low | High |
|---|---:|---:|
| Manual setup hours | 24 | 60 |
| AI setup hours, including review | 15.6 | 54 |
| Manual monthly admin hours | 6 | 16 |
| AI monthly admin hours, including review | 4.8 | 16 |
| Manual 36 month total | $25,436 | $52,400 |
| Claude Max 5x subscription, 36 months | $23,823 | $53,891 |
| ChatGPT Pro 100 subscription, 36 months | $23,823 | $53,891 |
| Claude Opus 5.5 API setup / monthly fee | $3.82 / $1.91 | $30 / $12 |
| Claude Opus 5.5 API, 36 months | $21,995 | $52,453 |
| GPT-6 Astra API setup / monthly fee | $9.85 / $4.925 | $75 / $30 |
| GPT-6 Astra API, 36 months | $22,110 | $53,146 |

Manual total = S x $68.09 + 36 x (H + A x $68.09). AI total = S x q x $68.09 + T0 + 36 x (H + A x r x $68.09 + Tm). S/A are manual setup/monthly hours, q/r the AI shares, H = $252.6255 monthly hosting plus backup, and T0/Tm setup/monthly tool fees. We round totals only after calculating.

We charge one $100 setup month for either subscription, then allocate $50 monthly for 36 operating months only when other work uses the remaining half. With exclusive use, add $1,800 to either subscription total. Equal prices do not mean equal allowances; check limits before work starts. [42, 43]

API fees use our cumulative IT budgets: setup input/output of 1 million/100,000 tokens low or 5 million/500,000 high; monthly input/output of 500,000/50,000 low or 2 million/200,000 high. Low input is 30% ordinary, 10% short cache writes and 60% reads; high input is uncached. Retries and billed reasoning are included; these are separate metered charges, with no subscription credits deducted. [44, 45]

Claude Code or Codex can prepare Texera's Compose configuration and restore commands for IT review. Desktop Commander on the authorized server can install dependencies, edit configuration and read container logs; its free local service is included here. Optional Remote Pro adds $20 for setup and $10 monthly at the same allocation, or $380 total. [46, 47, 48]

The Claude in Chrome extension can inspect the provider panel, Cloudflare DNS and Texera workflow UI while you watch. ChatGPT Work offers a separate cloud browser where available; confirm account access rather than promise an equivalent Chrome extension. [49, 50]

Use a new or staging server with approved commands or a restricted sandbox, and keep secrets out of chat. Use subscriptions for supervised supported tools and API keys for custom automated pipelines. Review changes, watch dashboard actions and test a PostgreSQL and workflow-file restore; we still own backups, updates and the decision. [51, 52]

Our conclusion stays unchanged: Deepnote's $2,808 base subscription remains cheaper. In the high case, recurring tool charges exceed setup savings within 36 months because monthly labor does not fall.


API worksheet: tool fee = (ordinary input x input rate + cache writes x write rate + cache reads x read rate + output x output rate) / 1,000,000. Rates per million tokens, ordered input/write/read/output: Claude Opus 5.5 = $4/$5/$0.20/$20; GPT-6 Astra = $10/$12.50/$1/$50. Use standard synchronous short-context rates; sum calls across sessions, including retries. A cache write replaces ordinary input billing. Keep application model usage separate from this setup-agent budget. Our assumptions exclude bespoke development, data cleaning and model training, as in the manual worksheet.

See the [post's setup explanation](https://jwatte.com/blog/apache-ai-machine-learning-small-business/#setup-time-when-claude-or-chatgpt-does-the-typing) for the comparison and limits.

### OpenNLP versus one Comprehend standard analysis

One Small server, 4 to 12 setup hours and 1 to 3 monthly hours, fixed 100 GB backups. [Comprehend](https://aws.amazon.com/comprehend/pricing/): fixed 10,000 texts of 550 characters each: 10,000 x ceiling(550/100) units x $0.0001 = $6/month, one standard API operation, three-unit minimum per request, ongoing list pricing without introductory free tier. Custom categories, training and multiple analyses need different costing.

| Cost | Low labor case | High labor case |
|---|---:|---:|
| Setup | $272 | $817 |
| Monthly operations | $93 | $229 |
| Self hosted, 36 months | $3,610 | $9,057 |
| Alternative monthly | $6 | $6 |
| Alternative, 36 months | $216 | $216 |
| Savings from self hosting | Negative $3,394 | Negative $8,841 |
| Break even month | Never | Never |

### Burr versus a conditional Zapier starting plan

One Small coordinator, 4 to 12 setup hours and 1 to 3 monthly hours, fixed 100 GB backups. No model calls. [Zapier Professional](https://zapier.com/pricing) starts at $19.99/month billed yearly, $239.88/year. Confirm task allowance and required features; the checked card did not bind the starting price to a task count. This is a subscription-floor comparison, not a verified quote for either sample workflow.

| Cost | Low labor case | High labor case |
|---|---:|---:|
| Setup | $272 | $817 |
| Monthly operations | $93 | $229 |
| Self hosted, 36 months | $3,610 | $9,057 |
| Alternative monthly | $20 | $20 |
| Alternative, 36 months | $720 | $720 |
| Savings from self hosting | Negative $2,890 | Negative $8,338 |
| Break even month | Never | Never |

## Sources

Read October 29, 2026. Review [ASF security](https://www.apache.org/security/) and [Apache License terms](https://www.apache.org/licenses/LICENSE-2.0); model and dataset terms can differ. Before clinical-text use of cTAKES, check HIPAA obligations. No privacy, medical, legal or accuracy assurance is made.

1. [Installing Apache Texera using Docker, Apache Texera](https://texera.incubator.apache.org/docs/v1.2.0/getting-started/installing-using-docker/). Read October 29, 2026.
2. [Apache Texera 1.2.0-incubating, Apache Texera](https://texera.incubator.apache.org/releases/1.2.0-incubating/). Read October 29, 2026.
3. [Apache OpenNLP Developer Documentation](https://opennlp.apache.org/docs/2.5.12/manual/opennlp.html). Read October 29, 2026.
4. [Download, Apache OpenNLP](https://opennlp.apache.org/download.html). Read October 29, 2026.
5. [Security, Apache OpenNLP](https://opennlp.apache.org/security.html). Read October 29, 2026.
6. [Installation, Apache Burr](https://burr.incubator.apache.org/docs/getting_started/install/). Read October 29, 2026.
7. [Human in the Loop, Apache Burr](https://burr.incubator.apache.org/docs/concepts/human-in-the-loop/). Read October 29, 2026.
8. [Deployment, Apache Burr](https://burr.incubator.apache.org/docs/ui/deployment/). Read October 29, 2026.
9. [Home, Apache Mahout](https://mahout.apache.org/). Read October 29, 2026.
10. [Incubation Policy](https://incubator.apache.org/policy/incubation.html). Read October 29, 2026.
11. [mahout release mahout-qumat-0.6.0 on GitHub](https://github.com/apache/mahout/releases/tag/mahout-qumat-0.6.0). Read October 29, 2026.
12. [Apache Burr (Incubating), Build Reliable AI Agents and Applications](https://burr.incubator.apache.org/). Read October 29, 2026.
13. [burr release v0.43.0-incubating on GitHub](https://github.com/apache/burr/releases/tag/v0.43.0-incubating). Read October 29, 2026.
14. [apache/ctakes on GitHub](https://github.com/apache/ctakes). Read October 29, 2026.
15. [projects.apache.org, projects.json](https://projects.apache.org/json/foundation/projects.json). Read October 29, 2026.
16. [apache/magpie on GitHub](https://github.com/apache/magpie). Read October 29, 2026.
17. [magpie release 0.1.0 on GitHub](https://github.com/apache/magpie/releases/tag/0.1.0). Read October 29, 2026.
18. [Apache Maka (Incubating)](https://maka.incubator.apache.org/). Read October 29, 2026.
19. [Index of /incubator/maka/0.2.0-incubating](https://downloads.apache.org/incubator/maka/0.2.0-incubating/). Read October 29, 2026.
20. [Apache OpenNLP](https://opennlp.apache.org). Read October 29, 2026.
21. [News, Apache OpenNLP](https://opennlp.apache.org/news/). Read October 29, 2026.
22. [Apache SINGA · Distributed deep learning system](https://singa.apache.org). Read October 29, 2026.
23. [Index of /singa/5.1.0](https://downloads.apache.org/singa/5.1.0/). Read October 29, 2026.
24. [Apache Sourcelume](https://sourcelume.apache.org). Read October 29, 2026.
25. [Apache SystemDS, An open source ML system for the end-to-end data science lifecycle](https://systemds.apache.org). Read October 29, 2026.
26. [Index of /systemds/3.4.0](https://downloads.apache.org/systemds/3.4.0/). Read October 29, 2026.
27. [Apache Texera](https://texera.incubator.apache.org/). Read October 29, 2026.
28. [texera release v1.2.0-incubating on GitHub](https://github.com/apache/texera/releases/tag/v1.2.0-incubating). Read October 29, 2026.
29. [Apache TVM](https://tvm.apache.org/). Read October 29, 2026.
30. [tvm release v0.27.0.post1 on GitHub](https://github.com/apache/tvm/releases/tag/v0.27.0.post1). Read October 29, 2026.
31. [Apache UIMA, Apache UIMA](https://uima.apache.org). Read October 29, 2026.
32. [Apache UIMA, Downloads](https://uima.apache.org/downloads.cgi). Read October 29, 2026.
33. [Occupational Employment and Wage Statistics Profiles](https://data.bls.gov/oesprofile/?areas=INDUSTRY,STATE,MSA&major_group=150000&measure=01&occupation=151244&year=2025). Read October 29, 2026.
34. [Bureau of Labor Statistics Data](https://data.bls.gov/timeseries/CMU2020000000000P). Read October 29, 2026.
35. [Droplet Pricing, DigitalOcean](https://www.digitalocean.com/pricing/droplets). Read October 29, 2026.
36. [B2 Cloud Storage Pricing, Backblaze](https://www.backblaze.com/cloud-storage/pricing). Read October 29, 2026.
37. [Deepnote Pricing](https://deepnote.com/pricing). Read October 29, 2026.
38. [Amazon Comprehend Pricing](https://aws.amazon.com/comprehend/pricing/). Read October 29, 2026.
39. [Plans & Pricing, Zapier](https://zapier.com/pricing). Read October 29, 2026.
40. [ASF Security Team, Apache Software Foundation](https://www.apache.org/security/). Read October 29, 2026.
41. [apache.org, LICENSE 2.0](https://www.apache.org/licenses/LICENSE-2.0). Read October 29, 2026.

42. [Claude Max plans](https://support.claude.com/en/articles/11049741-what-is-the-max-plan). Observed October 29, 2026, 23:31:40 UTC; supplied research.
43. [ChatGPT pricing](https://chatgpt.com/pricing/). Observed October 29, 2026, about 23:27 UTC; supplied research.
44. [Claude API pricing](https://platform.claude.com/docs/en/about-claude/pricing). Observed October 29, 2026, 23:31:14 UTC; supplied research.
45. [OpenAI API pricing](https://developers.openai.com/api/docs/pricing). Observed October 29, 2026, 23:31:13 UTC; supplied research.
46. [Codex CLI](https://developers.openai.com/codex/cli). Observed October 29, 2026, 23:32:01 UTC; supplied research.
47. [Desktop Commander pricing](https://desktopcommander.app/pricing/). Observed October 29, 2026, 23:31:08 UTC; supplied research.
48. [Desktop Commander documentation](https://raw.githubusercontent.com/wonderwhy-er/DesktopCommanderMCP/main/README.md). Observed October 29, 2026, 23:31:41 UTC; supplied research.
49. [Claude in Chrome](https://support.claude.com/en/articles/12012173-get-started-with-claude-in-chrome). Observed October 29, 2026, 23:31:11 UTC; supplied research.
50. [ChatGPT Work cloud browser](https://help.openai.com/en/articles/20001280-using-cloud-browser-in-chatgpt). Observed October 29, 2026, 23:32:01 UTC; supplied research.
51. [Anthropic Consumer Terms](https://www.anthropic.com/legal/consumer-terms). Observed October 29, 2026, 23:31:08 UTC; supplied research.
52. [OpenAI Terms of Use](https://openai.com/policies/terms-of-use/). Observed October 29, 2026, about 23:27 UTC; supplied research.
53. [METR experienced developer study](https://metr.org/blog/2025-07-10-early-2025-ai-experienced-os-dev-study/). Observed October 29, 2026, 23:30:58 UTC; supplied research.
54. [Codex with a ChatGPT plan](https://help.openai.com/en/articles/11369540-using-codex-with-your-chatgpt-plan). Observed October 29, 2026, 23:31:16 UTC; supplied research.

*Apache, the Apache feather logo and the Apache project names are trademarks of The Apache Software Foundation. This site is not affiliated with or endorsed by the ASF. Other product names are used for identification only.*