Odysseus Issue Taxonomization and Graph #4084

SynapticSmith · 2026-06-12T15:55:31Z

SynapticSmith
Jun 12, 2026

[✔️] Mostly AI Generated

To help you manage the 576 open issues for the Odysseus project, I have analyzed the provided list alongside the codebase architecture. By aligning the reported issues with the internal directory structure (e.g., services/hwfit/ for Cookbook, src/agent_tools for agent capabilities, and routes/ for UI endpoints), we can map these issues into a structured taxonomy.

Below is the comprehensive organization document. Potential duplicates have been consolidated, and all issue numbers are now directly linked to the repository for seamless navigation.

📊 Statistical Summary of Analyzed Issues

Based on the provided dataset of open issues, here is the distribution across the primary domains representing the project's current operational bottlenecks:

Domain	Concentration	Primary Bottlenecks
UI/UX & Frontend	~29%	Mobile responsiveness and component scaling
Agent Engine & Core Logic	~25%	Tool gating and context window amnesia
Model Providers & Hardware	~21%	GPU detection and Windows/Ollama parity
External Integrations	~15%	Calendar/Email timezone bugs and Search APIs
Infrastructure & Security	~10%	CI/CD pipelines and macOS/Windows build scripts

🗂️ Odysseus Issue Taxonomy

1. 🖥️ UI/UX & Frontend Platform (`static/js/`, `static/css/`)

Accessibility & Responsiveness

#3467 UI SCALE OPTIMIZATION
#3969 Cannot insert newlines in chat on mobile devices
#3962 The edit textbox for user chats is too narrow
#3203 iOS PWA layout does not fully respect safe areas
#977 Font size is too large / illegible on low-resolution screens
#1074 Panel docking only supports left and bottom snap zones
#3497 mobile: can't access top controls on pwa mode on iPhone

Navigation & Layout

#4071 codebase readability improvements for contributor onboarding
#3860 feat(ui): reorder sidebar tools and icon rail via drag-and-drop
#3871 fix(ui): icon rail and full sidebar both visible after expanding
#3123 Sidebar collapses when opening Calendar but not other tools
#3789 Add tabs for multiple windows snapped to the same side
#2245 Add a hamburger menu for the landing page

Components & State Management

#3878 Chat User Message Scroll Marker
#4045 Toast notifications disappear too quickly to read (1200ms default)
#3893 Drop downs hang over card edge on Scan/Download cookbook card
#4001, #4002, #3547, #3546 Brain/skills editing box bugs and editor mode
#3993 Live chat leaves executed email tool fences visible until reload
#4035 window rendering error when email view is maximized

Theming & Customization

#3692 Add default themes based on the catppuccin palettes
#3679 Optional animated login theme (admin toggle)
#3197 Custom background images for odysseus

2. 🧠 Agent Engine & Core Logic (`src/agent_tools/`, `src/chat_processor.py`)

Reasoning & Tool Execution

#3041 Agent Mode Is Broken and It barely perform as an agent
#3998 Agent mode: next round's reasoning streams into the reply bubble
#3992 Agent persists pre-tool prose before the real tool result
#3604 Agent mode executes ordinary Markdown code fences as tool calls
#3668 Agent intent-without-action supervisor (_INTENT_RE) is English-only
#3792 Task Scheduler: filter thinking deltas from agent loop results
#3436 Need HITL (Human-in-the-loop) in stream_agent_loop to prevent infinite error loops

Context & Memory Management

#4050 add memory pruning, and corresponding "deep memory" search for long persistence
#3890, #3889, #3888 Memory tidy lifecycle, OAuth credentials, and 100-entry limit
#3729 Native TRACE-inspired and MemMachine Hierarchical Memory Engine
#3745 Memory Amnesia Bug Fix & Subsystem Hardening (Context Compactor + CRUD Tools)
#2750 Agent prompt token bloat: measure, slim, and modularize
#1893 Hierarchical B+Tree memory for Agent mode to fix context bloat

Tool Access & Routing Filters

#4048 Agent without ChromaDB has NO file tools
#3766 _classify_agent_request() strips all tools for non-English queries
#3934 Tool-RAG low-signal gate hides most available tools
#3709 Opt-in hard gate: block high-impact tools when the agent's context is untrusted
#3605 Agent never offered generate_image/edit_image

3. ⚙️ Model Providers & Hardware / Cookbook (`services/hwfit/`)

Local Providers & Hardware Detection

#4059, #4017, #3995, #3897 Cookbook downloader crashes, silent stalls, and orphaned hf/python processes
#4049 Model files disappear after update/rebuild (cache unawareness)
#3882 detect_system() no-GPU result missing gpus/gpu_groups/homogeneous keys
#3939 Scanning for models appears to spike CPU usage
#1504, #488, #528 Llama.cpp ROCm support and AMD GPU misclassification
#4056 Cookbook Serve: per-model compute backend selector
#3624 Add "wsl" as a cookbook server platform

External APIs & Remote Integrations

#4037 feat: Add native Google Gemini API provider support
#3145 Gemini-native models fall back to OpenAI-compatible chat/completions
#3021 Add Perplexity as an OpenAI-compatible provider
#2462 Add NEAR AI as a first-class provider (confidential/TEE inference)
#13 feat: AWS Bedrock integration

4. 🔌 External Integrations (`services/search/`, `routes/`)

Email & Calendar (CalDAV/IMAP)

#3999 feat(email): Microsoft Graph (Exchange/Outlook) mail accounts and agent access
#3915, #3832, #3631 Timezone bugs (daily_brief, Reminders, and Calendar week view defaulting to UTC)
#4058 manage_calendar list_events returns no events when start == end
#3869 Email Calendar Events CalDAV Calendar Selection and Migration
#3345 Support Proton Mail accounts via Proton Mail Bridge

Web Search & Research

#3344 feat(search): add Perplexity Search as a selectable web-search provider
#4055 Bug: Polish-language web search requests classified as "low-signal"
#3696 Migrate to the maintained ddgs package (duckduckgo-search is frozen)
#2866 Add SerpApi as a search provider
#2703 Kagi - add as a search provider
#531 Add Exa as an additional search provider

MCP (Model Context Protocol) & Webhooks

#4026 Built-in Browser MCP (@playwright/mcp) reports "not installed in npx cache"
#3877 MCP Streamable HTTP client crashes with "Attempted to exit cancel scope in a different task"
#2976, #2975, #2974 Add MS Teams, Slack, and Mattermost Webhook presets

5. 🏗️ Infrastructure, DevOps & Security (`.github/workflows/`, `core/`)

Testing & CI/CD Pipelines

#3983 tests: plan oversized test-file splits
#3973 tests: add non-blocking order-sensitivity reporting
#3968 install: Fresh install and smoke-test coverage tracker
#3694 Stabilization: slow down risky merges with clear PR review rules

Security & Data Hardening

#3803 Hardening audit: unversioned migrations, non-atomic writes, PII in logs, missing owner scoping
#3866 docs/security: add a minimal accountability checklist for privileged agent actions
#3966 API token PATCH should reject non-object JSON bodies instead of 500ing
#3200 DNS Rebinding SSRF in webpage content fetcher allows access to private networks

Platforms & Packaging (Docker, Windows, macOS)

#3896 Parameterize Docker volume paths for external orchestrators
#3846 Docker on Windows: entrypoint re-chowns multi-GB caches every boot
#3815 build-macos-app.sh produces unsigned bundle that Gatekeeper silently kills
#3747 Refactor launch-windows.ps1 for uv installation and Python management

6. 📚 Documents, Notes & Canvas (`src/document_processor.py`)

PDF & Formatting Engine

#3861 Typst Support
#3447 PDF form: unchecked AcroForm checkboxes render and export as checked
#3327 PDF OCR fails silently for image-only/scanned PDFs
#2135 PDF form export drops fields with non-ASCII names

Editor & Notes Panel

#4066 Notes: per-item and whole-note agent integration
#4065 Notes: per-note sharing (view/edit collaborators) with its security hardening
#4064 Notes: preview panel as the single editor
#3788 Notes pane renders behind other views and chat controls
#2909 Proper DOCX editing in Documents

Library & Attachments

#3938 Proper document library
#3750 Make individual documents from the library drag&drop elements
#3646 Add ability to attach files directly from the library

7. 🗄️ Workspace, Research & Memory (`services/research/`)

Deep Research Pipeline

#2787 [feat] Deep Research: allow manual search provider override
#2339 Add an expand/fullscreen control to the Deep Research synapse graph
#2741 Add backup support for Deep Research runs
#2718 The Agent can trigger the deep reasearch but cant draw context from the result

Knowledge & Extraction

#3965 Skill extraction can save the later JSON object from multi-object output
#3854 Visual report title extraction deletes a mid-document section heading

Session & History Management

#4005 Research "Discuss" (spin-off) chats lose the report under compaction
#3531 Backspace while renaming a chat opens the delete-session prompt
#3402 feat(library): show cumulative session token count in chat cards

8. 🔐 Authentication & Permissions (`core/auth.py`)

Access Control & Isolation

#4061 Notes routes treat an identity-less request as the single-user mode (fail-open)
#4052 Bearer/API-token clients are siloed as the "api" pseudo-user on chat, model, and upload routes
#3149 Add default-deny API token capability checks for bearer routes

Login & Auth Providers

#3561 Login returns 500 when password is longer than 72 bytes (bcrypt ValueError)
#2960 Design a reusable OAuth/device-flow provider component
#806 SSO via OIDC support

9. 🛠️ Tooling, Scripts & Subprocesses (`scripts/`)

CLI & Execution Environment

#3446 shell exec: timeout=0 times out immediately instead of running unbounded
#3430 Agent bash/python tool output only appears after command finishes
#2957 Windows agent shell tool: stray quote breaks commands + defaults to Unix commands
#2759 feat: add native PowerShell agent tool alongside bash

Diagnostics & Dev Tools

#3060 feat: Trace collector dev-tool for flagging successful/failed agent interactions
#2122 odysseus-* CLI list/search commands return wrong or missing results

🛡️ Architectural Resolution Strategy

Based on the THREAT_MODEL.md and README.md documents provided in the codebase, many of these issues intersect directly with core security and architectural rules. When tackling this backlog, apply these strict codebase invariants:

Enforce Owner-Scoping: A major pattern of bugs (e.g., #4052, #2352, #2286) stems from bypassing or mishandling the owner_filter logic. As stated in the threat model, "A caller must NEVER see another owner's row". All CRUD operations affecting memory, tokens, tasks, and documents must pass through get_current_user or the bearer-token owner-attribution sentinel before hitting the database.
Un-Gate the Agent Safely: Issues requesting more open tool execution (#3604, #3605) must be balanced against src/tool_security.py. High-risk capabilities (shell, Python execution, file writes) must remain admin-only by default (NON_ADMIN_BLOCKED_TOOLS).
Sanitize Untrusted Data: Issues dealing with Deep Research ingestion or webpage parsing (#2135, #3854) must be routed through untrusted_context_message() to prevent prompt-injection attacks from poisoning the RAG pipeline.

🗺️ Odysseus Issue Taxonomy Graph (Visual Map)

This text-based tree structure maps out the categories defined above. You can directly copy it into the repository's ROADMAP.md or CONTRIBUTING.md to guide new contributors.

Odysseus Project (576 Open Issues)
│
├── 1. 🖥️ UI/UX & Frontend Platform
│   ├── Accessibility & Responsiveness (Mobile PWA, UI Scaling)
│   ├── Navigation & Layout (Sidebar, Multi-window tabs)
│   ├── Components & State (Chat scroll, Modals, Toasts)
│   └── Theming & Customization (Catppuccin, Dark Mode)
│
├── 2. 🧠 Agent Engine & Core Logic
│   ├── Reasoning & Tool Execution (Markdown parsing, infinite loops)
│   ├── Context & Memory Management (Amnesia, Context Compaction)
│   └── Tool Access & Routing Filters (Language gating, Untrusted Context)
│
├── 3. ⚙️ Model Providers & Hardware (Cookbook)
│   ├── Local Providers & Hardware (AMD ROCm, Windows/WSL, llama.cpp)
│   └── External APIs (Gemini, Perplexity, Bedrock)
│
├── 4. 🔌 External Integrations
│   ├── Email & Calendar (Outlook, IMAP, CalDAV, Timezones)
│   ├── Web Search (DuckDuckGo, SerpAPI, Perplexity)
│   └── MCP & Webhooks (Playwright MCP, Slack/Teams)
│
├── 5. 🏗️ Infrastructure, DevOps & Security
│   ├── Testing & CI/CD Pipelines (Test splits, PR rules)
│   ├── Security & Data Hardening (SSRF, Auth boundaries)
│   └── Platforms & Packaging (Docker multi-stage, macOS/Windows native)
│
├── 6. 📚 Documents, Notes & Canvas
│   ├── PDF & Formatting Engine (OCR, Typst, AcroForm)
│   ├── Editor & Notes Panel (Markdown unified model, Collab)
│   └── Library & Attachments (Drag & Drop, Zip export)
│
├── 7. 🗄️ Workspace, Research & Memory
│   ├── Deep Research Pipeline (Fallback providers, Synapse Graph)
│   └── Session & History Management (Discuss spin-offs, DB migrations)
│
├── 8. 🔐 Authentication & Permissions
│   ├── Access Control (Bearer tokens, API pseudo-user)
│   └── Login & Auth Providers (OIDC, OAuth2, Rate Limits)
│
└── 9. 🛠️ Tooling, Scripts & Subprocesses
    ├── CLI & Execution Environment (PowerShell tool, Bash timeouts)
    └── Diagnostics & Dev Tools (Trace collector)

🏷️ Recommended GitHub Label Taxonomy

To map this taxonomy directly into GitHub, I recommend creating a standardized labeling system to bulk-apply and instantly organize the repository.

Domain Labels (Color: #1d76db - Blue):

domain: ui/ux
domain: agent-engine
domain: hardware/cookbook
domain: integrations
domain: memory/context
domain: infrastructure/ci

Type Labels (Color: #e99695 - Red/Pink):

type: bug (Core logic failures, crashes)
type: enhancement (New features)
type: papercut (Minor UI annoyances, typos)
type: security (Hardening, SSRF, prompt injection)
type: developer-experience (Build scripts, docs, testing)

Platform Labels (Color: #0e8a16 - Green):

platform: windows
platform: macos
platform: linux/docker
platform: mobile/pwa

Status/Triage Labels (Color: #fbca04 - Yellow):

status: needs-validation
status: blocked
status: good-first-issue

🚀 Recommendations & Phased Execution Plan

Consolidate Redundancies: You have highly related clusters in "Timezone/UTC Calendar bugs" (#3915, #3832, #3631) and "Cookbook Download Stalls" (#4059, #4017, #3995). Creating GitHub Epic tracker issues for these will collapse them and simplify the backlog.

With 576 issues, a maintainer cannot tackle everything at once. Use this 4-phase approach to burn down the backlog systematically:

Phase 1: Stop the Bleeding (P0/P1 Security & Blockers)

Focus: Security vulnerabilities and critical crashes.
Target Issues: #3200 (DNS Rebinding SSRF), #2891 (RCE via manage_mcp), #1940 (Memory lost-update race condition), and Cookbook download crashes (#4059, #2722).
Action: Lock down the untrusted context boundaries, secure the bearer tokens (#3149), and fix the concurrent write locks on the SQLite/JSON memory files.

Phase 2: Agent Stability & Core User Experience

Focus: Making sure the AI actually behaves like an agent.
Target Issues: Un-gating the RAG tools (#4048, #3766, #3934), fixing context amnesia (#3745), and preventing markdown code-fence execution loops (#3604).
Action: Rewrite the _classify_agent_request to be less aggressive and language-agnostic. Implement strict HITL (Human-in-the-loop) circuit breakers for infinite loops.

Phase 3: Hardware Parity & Provider Integrations

Focus: Ensuring users can actually run the local models.
Target Issues: AMD ROCm detection (#1504, #488), Windows path quoting (#2957, #2642), and resolving API integration timezone bugs (#3915, #3832).
Action: Consolidate the platform-specific hardware checks into a unified hwfit module. Standardize all datetime parsing to strictly capture and respect the user's local timezone offset over UTC.

Phase 4: UX Papercuts & "Good First Issues"

Focus: Quality of life and onboarding new contributors.
Target Issues: UI scaling (#3467), edit textbox sizes (#3962), missing hamburger menus (#2245), and drag-and-drop support (#3750).
Action: Create a GitHub Project Board specifically for "Frontend Papercuts" and tag them all as good-first-issue to attract open-source contributors to help clear out the remaining ~30% of the backlog.

🤖 Automation: Auto-Labeling Script

Since manually organizing and labeling 576 issues is a massive undertaking, the most practical next step is automation.

If you have maintainer access, you can use the GitHub API to bootstrap this organization. Save the following Python script using the PyGithub library as scripts/triage_bot.py. It will automatically create the taxonomy labels defined above and begin categorizing open issues based on title and body heuristics.

(Note: You will need a GitHub Personal Access Token assigned to GITHUB_TOKEN with repository write access).

import os
from github import Github

# Configuration
REPO_NAME = "pewdiepie-archdaemon/odysseus"
TOKEN = os.getenv("GITHUB_TOKEN")

# Define our new Taxonomy Labels (Name, Color, Description)
LABELS = {
    # Domains
    "domain: ui/ux": ("1d76db", "Frontend, styling, mobile, and accessibility"),
    "domain: agent-engine": ("1d76db", "Core LLM routing, RAG gates, and reasoning loops"),
    "domain: hardware/cookbook": ("1d76db", "Local model serving, GPUs, Ollama, vLLM"),
    "domain: integrations": ("1d76db", "Email, Calendar, Web Search, MCP"),
    "domain: memory/context": ("1d76db", "Context window, ChromaDB, vector storage"),
    "domain: infrastructure/ci": ("1d76db", "Docker, build scripts, GitHub Actions"),
    
    # Types
    "type: bug": ("e99695", "Something isn't working correctly"),
    "type: enhancement": ("a2eeef", "New feature or request"),
    "type: security": ("d73a4a", "Vulnerabilities, SSRF, auth bypass"),
    
    # Platforms
    "platform: windows": ("0e8a16", "Windows native or WSL specific"),
    "platform: macos": ("0e8a16", "macOS or Apple Silicon specific"),
    "platform: docker": ("0e8a16", "Containerized deployment specific")
}

# Simple keyword heuristics for auto-tagging
KEYWORD_MAP = {
    "domain: ui/ux": ["ui", "frontend", "mobile", "pwa", "css", "scroll", "button", "sidebar"],
    "domain: hardware/cookbook": ["cookbook", "rocm", "amd", "gpu", "cuda", "vllm", "llama.cpp", "gguf"],
    "domain: integrations": ["email", "calendar", "outlook", "imap", "mcp", "searxng", "duckduckgo"],
    "domain: memory/context": ["memory", "chromadb", "rag", "context", "amnesia"],
    "platform: windows": ["windows", "ps1", "powershell", ".bat"],
    "platform: macos": ["macos", "apple silicon", "m1", "m2", "m3", "gatekeeper"]
}

def setup_labels(repo):
    print("Setting up taxonomy labels...")
    existing_labels = {l.name: l for l in repo.get_labels()}
    
    for name, (color, desc) in LABELS.items():
        if name in existing_labels:
            print(f"Updating label: {name}")
            existing_labels[name].edit(name=name, color=color, description=desc)
        else:
            print(f"Creating label: {name}")
            repo.create_label(name=name, color=color, description=desc)

def auto_triage_issues(repo):
    print("Scanning open issues for auto-triage...")
    issues = repo.get_issues(state='open')
    
    for issue in issues:
        if issue.pull_request: # Skip PRs
            continue
            
        text_to_search = (str(issue.title) + " " + str(issue.body)).lower()
        tags_to_apply = []
        
        for label, keywords in KEYWORD_MAP.items():
            if any(kw in text_to_search for kw in keywords):
                tags_to_apply.append(label)
                
        if tags_to_apply:
            print(f"Tagging #{issue.number} ({issue.title[:30]}...) with {tags_to_apply}")
            # issue.add_to_labels(*tags_to_apply) # UNCOMMENT TO ACTUALLY APPLY

if __name__ == "__main__":
    if not TOKEN:
        print("Please set GITHUB_TOKEN environment variable.")
        exit(1)
        
    g = Github(TOKEN)
    repo = g.get_repo(REPO_NAME)
    
    setup_labels(repo)
    # auto_triage_issues(repo) # Run cautiously!
    print("Done.")

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Odysseus Issue Taxonomization and Graph #4084

Uh oh!

{{title}}

Uh oh!

Uh oh!

{{editor}}'s edit

{{editor}}'s edit

Uh oh!

Replies: 0 comments

Select a reply

Uh oh!

Odysseus Issue Taxonomization and Graph #4084

Uh oh!

Uh oh!

SynapticSmith Jun 12, 2026

📊 Statistical Summary of Analyzed Issues

🗂️ Odysseus Issue Taxonomy

1. 🖥️ UI/UX & Frontend Platform (static/js/, static/css/)

2. 🧠 Agent Engine & Core Logic (src/agent_tools/, src/chat_processor.py)

3. ⚙️ Model Providers & Hardware / Cookbook (services/hwfit/)

4. 🔌 External Integrations (services/search/, routes/)

5. 🏗️ Infrastructure, DevOps & Security (.github/workflows/, core/)

6. 📚 Documents, Notes & Canvas (src/document_processor.py)

7. 🗄️ Workspace, Research & Memory (services/research/)

8. 🔐 Authentication & Permissions (core/auth.py)

9. 🛠️ Tooling, Scripts & Subprocesses (scripts/)

🛡️ Architectural Resolution Strategy

🗺️ Odysseus Issue Taxonomy Graph (Visual Map)

🏷️ Recommended GitHub Label Taxonomy

🚀 Recommendations & Phased Execution Plan

🤖 Automation: Auto-Labeling Script

Replies: 0 comments

SynapticSmith
Jun 12, 2026

1. 🖥️ UI/UX & Frontend Platform (`static/js/`, `static/css/`)

2. 🧠 Agent Engine & Core Logic (`src/agent_tools/`, `src/chat_processor.py`)

3. ⚙️ Model Providers & Hardware / Cookbook (`services/hwfit/`)

4. 🔌 External Integrations (`services/search/`, `routes/`)

5. 🏗️ Infrastructure, DevOps & Security (`.github/workflows/`, `core/`)

6. 📚 Documents, Notes & Canvas (`src/document_processor.py`)

7. 🗄️ Workspace, Research & Memory (`services/research/`)

8. 🔐 Authentication & Permissions (`core/auth.py`)

9. 🛠️ Tooling, Scripts & Subprocesses (`scripts/`)