February 6, 2026 5 min read Python fork

awesome llm apps

Abstract

Collection of awesome LLM apps with AI Agents and RAG using OpenAI, Anthropic, Gemini and opensource models.

Forked from Shubhamsaboo/awesome-llm-apps

The Problem

This repository contains 7 self-contained LLM application projects that share no unified architecture, creating discoverability and integration challenges. Teams evaluating the stack must assess each project independently rather than understanding a coherent system. The collection spans autonomous game players, multi-agent teams, RAG tutorials, and voice agents—valuable individually but opaque as a portfolio.

What This Does

The repository organizes 1073 files across 7 distinct projects: advanced_ai_agents (462 files, 308 code files), ai_agent_framework_crash_course (250 files, 112 code files), advanced_llm_apps (174 files, 37 code files), rag_tutorials (72 files, 25 code files), starter_ai_agents (51 files, 20 code files), voice_ai_agents (12 files, 6 code files), and mcp_ai_agents (17 files, 5 code files). Recurring techniques include RAG implementations (rag_tutorials/ai_blog_search/app.py, rag_tutorials/rag_chain/app.py), multi-agent team patterns (advanced_ai_agents/multi_agent_apps/agent_teams/ with 8 agent team variants), and autonomous agents for chess, tic-tac-toe, and 3D pygame. Entry points such as advanced_ai_agents/autonomous_game_playing_agent_apps/ai_tic_tac_toe_agent/app.py:34 reach 309 functions, while rag_tutorials/ai_blog_search/app.py:153 (agent) reaches 44 functions with no internal callers. The code touches 27 model inference calls, 41 filesystem operations, 76 database operations, and 27 outbound network calls across resolved edges.

How It Is Wired

Execution starts at individual entry points rather than a single root. The most connected module is advanced_ai_agents/multi_agent_apps/ai_news_and_podcast_agents/web/src/services/api (Ca 22, Ce 0, instability 0)—22 modules depend on it with high blast radius. The internal call graph has 1216 resolved call edges; execute_query is called from 95 places, Agent from 43, get_session from 26. Representative edges: main -> Agent x18, initialize_agents -> Agent x14, generate_travel_plan -> update_trip_plan_status x11. Five files carry the highest cognitive-load findings: Sidebar.js, ActivePodcastPreview.js, AudioConfirmation.js all have max indentation depth 10, making control flow hard to follow. Duplicated code blocks appear across 184 files with 3314 repeated 6-line patterns in legal and real estate agent teams—extract shared helpers to DRY the repeated logic. Two critical security findings exist: h11@0.14.0 CVE-2025-43859, gitpython@3.1.44 CVE-2026-42284, and ANTHROPIC_API_KEY committed in .github/workflows/claude.yml where contributor-controlled input reaches a job with publish tokens.

How To Use It

Setup: This is a collection of 7 projects with no single install command. Each project has its own requirements.txt. For example, advanced_ai_agents/autonomous_game_playing_agent_apps/ai_tic_tac_toe_agent/requirements.txt and advanced_ai_agents/multi_agent_apps/agent_teams/ai_travel_planner_agent_team/backend/requirements.txt. Clone with: git clone https://github.com/moses-y/awesome-llm-apps

Configuration: Environment variables are project-specific. The travel planner team uses .env.example at advanced_ai_agents/multi_agent_apps/agent_teams/ai_travel_planner_agent_team/backend/.env.example. The legal agent team has similar examples. No single .env or config file spans all projects.

Running it: Choose a project directory and invoke its entry point. To run the tic-tac-toe agent: python advanced_ai_agents/autonomous_game_playing_agent_apps/ai_tic_tac_toe_agent/app.py. For the RAG blog search: python rag_tutorials/ai_blog_search/app.py. Each project's README documents its specific startup command.

Real-World Use

A developer wants to evaluate multi-agent team patterns for a customer support system. They examine advanced_ai_agents/multi_agent_apps/agent_teams/ai_financial_coach_agent/ai_financial_coach_agent.py (13 functions, 13 classes, called from 1 other file) to understand agent role definitions, then look at advanced_ai_agents/multi_agent_apps/agent_teams/ai_legal_agent_team/legal_agent_team.py for team orchestration patterns. The ai_travel_planner_agent_team/backend/services/db_service.py (4 functions, called from 21 files) shows database service patterns with initialize_db_pool, close_db_pool, get_db_session, execute_query that could be adapted. The rag_tutorials/rag_chain/app.py:81 run_rag_chain reach 44 functions and demonstrates retrieval-augmented generation workflows.

Code Health & Issues

The static analysis found 189 total findings: 70 high, 119 medium, 0 low across 3 kinds. Specific findings:

[MEDIUM/clarity] Hub module: advanced_ai_agents/multi_agent_apps/ai_news_and_podcast_agents/web/src/services/api.js — 22 modules depend on this one; high-blast-radius churn. Keep stable and small; move volatile logic out.
[HIGH/cognitive_load] Deep nesting x58: advanced_ai_agents/multi_agent_apps/ai_news_and_podcast_agents/web/src/components/Sidebar.js, ActivePodcastPreview.js, AudioConfirmation.js — max indentation depth 10. Flatten with early returns/guard clauses; extract inner blocks.
[HIGH/clarity] Duplicated code blocks: advanced_ai_agents/multi_agent_apps/agent_teams/ai_legal_agent_team/legal_agent_team.py, local_ai_legal_agent_team/local_legal_agent.py, and real estate team variants — 3314 repeated 6-line blocks across 184 files. Extract shared helpers; DRY the repeated logic.

Code health audit (14 findings, 2 critical, 4 high, 6 medium, 2 low):

[CRITICAL] Upgrade the pinned dependency carrying a critical advisory: h11@0.14.0 CVE-2025-43859, gitpython@3.1.44 CVE-2026-42284, pyarrow@19.0.1 CVE-2026-25087. Fix: upgrade to fixed versions and commit lockfile.
[CRITICAL] Keep secrets out of workflows a fork can trigger: .github/workflows/claude.yml contains ANTHROPIC_API_KEY. Fix: move secret-using steps into workflow_run job that never checks out pull request code.
[HIGH] Pin third-party GitHub Actions to a commit SHA: anthropics/claude-code-action@beta in .github/workflows. Fix: replace @vN with 40-character commit SHA.
[HIGH] Upgrade the pinned dependency carrying a published advisory: protobuf@5.29.3 CVE-2026-0994, pyasn1@0.6.1 CVE-2026-23490, python-multipart@0.0.20 CVE-2026-53539, requests@2.32.3 CVE-2024-47081, and 10 more. Fix: upgrade and commit lockfile.
[HIGH] Commit a lockfile beside the manifest: advanced_ai_agents/multi_agent_apps/ai_news_and_podcast_agents/web/package.json has no lockfile. Fix: run package manager once and commit generated lockfile.
[HIGH] Make CI invoke the test suite it has: 11 test files, no test command in any workflow. Fix: add test step to existing workflow.
[MEDIUM] Enable Dependabot or Renovate: 143 manifests, no update bot configured. Fix: commit .github/dependabot.yml covering repo ecosystems plus github-actions.
[MEDIUM] Pin the container base image by digest: advanced_ai_agents/multi_agent_apps/agent_teams/ai_travel_planner_agent_team/backend/Dockerfile uses python:3.12-slim-bookworm (mutable). Fix: use python:3.12-slim-bookworm@sha256:<digest> and enable Dependabot docker ecosystem.

Beyond the measured findings: committed secrets in .github/workflows/claude.yml, 2 blobs over 5MB (streaming-ai-chatbot.gif 23.8MB, github thinkpath video.mp4 19.1MB) not on LFS, and no dependency vulnerability scanning in CI.

The Bottom Line

This repository is a valuable portfolio of 7 independently viable LLM application projects—game-playing agents, multi-agent teams, RAG systems, and voice interfaces—that demonstrate practical patterns across OpenAI, Anthropic, Gemini, and open-source models. However, the absence of a unified architecture, critical dependency vulnerabilities in pinned packages, secrets in workflows, and missing lockfiles mean it should be evaluated project-by-project rather than as an integrated system. Teams should clone the specific projects relevant to their needs, upgrade the called-out dependencies, and remove the committed API key before running any CI or sharing code.

What the analyser found

Deployment readiness

6/7

Container image
CI pipeline
Lockfile committed
Test suite
README
License
No committed secrets

Composition

1,073 files

Python417
Markdown205
JavaScript49
TSX26
TypeScript15
JSON14

ReactNext.jsFlaskExpressDockerLangChainVector store

Module dependencies

The 10 most depended-upon modules of 507, from static import analysis. Red outline marks a module in an import cycle.

graph LR n0__agents_web_src_services_api["api"] n1_agents_beifong_db_connection["connection"] n2_d_podcast_agents_web_src_App["App"] n3_s_web_src_components_Sidebar["Sidebar"] n4_nt_agent_ai_consultant_agent["ai_consultant_agent"] n5_nts_web_src_pages_StudioChat["StudioChat"] n6_ts_web_src_pages_SocialMedia["SocialMedia"] n7_rc_components_social_FeedTab["FeedTab"] n8_c_components_social_StatsTab["StatsTab"] n9_agents_web_src_pages_Voyager["Voyager"] n2_d_podcast_agents_web_src_App --> n5_nts_web_src_pages_StudioChat n2_d_podcast_agents_web_src_App --> n9_agents_web_src_pages_Voyager n2_d_podcast_agents_web_src_App --> n6_ts_web_src_pages_SocialMedia n3_s_web_src_components_Sidebar --> n0__agents_web_src_services_api n8_c_components_social_StatsTab --> n0__agents_web_src_services_api n6_ts_web_src_pages_SocialMedia --> n0__agents_web_src_services_api n6_ts_web_src_pages_SocialMedia --> n7_rc_components_social_FeedTab n6_ts_web_src_pages_SocialMedia --> n8_c_components_social_StatsTab n5_nts_web_src_pages_StudioChat --> n3_s_web_src_components_Sidebar n5_nts_web_src_pages_StudioChat --> n0__agents_web_src_services_api n9_agents_web_src_pages_Voyager --> n0__agents_web_src_services_api

Dependencies

21 of 81 declared dependencies are a major version behind. Resolved 81 of 81 against the registry.

Package	Declared	Latest	Behind
`rpds-py`	==0.23.1	2026.6.3	2026 major
`setuptools`	==75.8.0	84.0.0	9 major
`pyarrow`	==19.0.1	25.0.1	6 major
`websockets`	==14.2	17.0.1	3 major
`cachetools`	==5.5.2	7.1.7	2 major
`openai`	==1.64.0	3.1.0	2 major
`packaging`	==24.2	26.3	2 major
`protobuf`	==5.29.3	7.35.1	2 major

Findings

LowSecret-shaped paths present; the code-health audit confirms or clears them mcp_ai_agents/browser_mcp_agent/mcp_agent.secrets.yaml.example

Open in Code Brain → See its neighbourhood →

What would stop me shipping this

Ranked by severity × confidence × production reach. Reach is the honest discriminator across a collection that is mostly other people's code: the same finding matters more in something that ships.

critical2

high4

medium6

low2

Upgrade the pinned dependency carrying a critical advisorycritical30 occurrences

h11@0.14.0 CVE-2025-43859 (critical) +1 more; gitpython@3.1.44 CVE-2026-42284 (critical) +25 more; pyarrow@19.0.1 CVE-2026-25087 (critical) +1 more

The manifest pins the exact version the advisory covers, so this is not a warning about drift: the vulnerable code is what installs, and a critical rating means it is reachable over the network without credentials or user interaction.

Fix: Upgrade to the fixed version named in the advisory, then commit the lockfile.

Keep secrets out of workflows a fork can triggercritical

.github/workflows/claude.yml

ANTHROPIC_API_KEY

Contributor-controlled input reaches a job holding a publish token or cloud credential, so a pull request becomes credential exfiltration.

Fix: Move secret-using steps into a workflow_run job that never checks out pull request code, or gate on an environment with required reviewers.

Pin third-party GitHub Actions to a commit SHAhigh

.github/workflows

anthropics/claude-code-action@beta

A tag can be moved, so the action running with your token and secrets is whatever its owner last pushed; this is how tj-actions/changed-files leaked secrets from thousands of repos.

Fix: Replace each @vN with the 40-character commit SHA, keep # vN as a comment, and let Dependabot bump the SHAs.

Upgrade the pinned dependency carrying a published advisoryhigh89 occurrences

protobuf@5.29.3 CVE-2026-0994 (high) +3 more; pyasn1@0.6.1 CVE-2026-23490 (high) +9 more; python-multipart@0.0.20 CVE-2026-53539 (high) +13 more; requests@2.32.3 CVE-2024-47081 (high) +3 more; and 10 more package(s)

The pinned version is the one that installs, so a published advisory against it describes this deployment rather than a hypothetical one, and the advisory is also a public description of how to exploit it.

Fix: Upgrade to the fixed version named in the advisory, then commit the lockfile.

Commit a lockfile beside the manifesthigh

advanced_ai_agents/multi_agent_apps/ai_news_and_podcast_agents/web/package.json

manifest with no lockfile in scope

An unlocked range means the artifact you tested and the artifact you ship can contain different transitive code, so a malicious patch release reaches production with no diff.

Fix: Run the package manager once and commit the generated lockfile.

Make CI invoke the test suite it hashigh11 occurrences

.github/workflows

11 test files, no test command in any workflow

A green check that never executed an assertion is worse than no check, because reviewers trust it.

Fix: Add a test step to the existing workflow rather than a new workflow.

Enable Dependabot or Renovatemedium

143 manifest(s), no update bot configured

Without a bot a published advisory sits unpatched until someone audits by hand, which across 1,322 repositories means never.

Fix: Commit .github/dependabot.yml covering the repo ecosystems plus github-actions.

Pin the container base image by digestmedium2 occurrences

advanced_ai_agents/multi_agent_apps/agent_teams/ai_travel_planner_agent_team/backend/Dockerfile

python:3.12-slim-bookworm, python:3.12-slim-bookworm

An untagged or mutable base means today's build and last month's contain different libc and a different CVE set, with no record of which shipped.

Fix: Use image:tag@sha256:<digest> and enable Dependabot's docker ecosystem.

Gate pull requests on a dependency vulnerability scanmedium

.github/workflows

no dependency scan in CI

This is the one gate that would catch a known-vulnerable package before it reaches a build, and no repository in the sample had it.

Fix: Add dependency-review-action on pull_request, or osv-scanner on push and a schedule.

Move large binaries to Git LFS or out of the repositorymedium2 occurrences

advanced_llm_apps/chat_with_X_tutorials/streaming_ai_chatbot/docs/images/streaming-ai-chatbot.gif

2 blobs over 5MB: streaming-ai-chatbot.gif 23.8MB, github thinkpath video.mp4 19.1MB

One repository here carries twenty blobs over 5MB including a 10.9MB spreadsheet, so every clone and every CI checkout pays for data nobody diffs.

Fix: Track those extensions with LFS, or move datasets to object storage and fetch them in a setup step.

Add a non-root USER to the imagemedium

advanced_ai_agents/multi_agent_apps/agent_teams/ai_travel_planner_agent_team/backend/Dockerfile

CMD or ENTRYPOINT with no USER directive

A process running as root in the container is root against every mounted volume, and it turns any container escape or writable-mount mistake from a contained problem into a host one.

Fix: Create an unprivileged user, chown what it needs, and end the Dockerfile with USER.

Expand the test suite; it cannot cover this codebasemedium9 occurrences

9 test files against 507 source files (ratio 0.018)

A reviewer reads "has tests" and assumes a safety net that covers two files out of eighty-five, which is how untested paths ship behind a green badge.

Fix: Add tests for the highest fan-in modules first.

2 further finding(s) are listed in the full report.

Checked deterministically against the repository tree and a bounded set of its files: committed credentials, unpinned actions and base images, missing lockfiles and update bots, workflows that discard failures, published advisories against the declared dependencies, runtime configuration, licensing and notebook reproducibility. No language model is involved in this section.

View on GitHub All Projects