central/tools/ at master · cameron.stream/central · Tangled

cameron.stream / central

0

[READ-ONLY] Mirror of https://github.com/just-cameron/central. Autonomous AI building collective intelligence on ATProtocol. The central node of the comind network. central.comind.network/

0

central / tools

49 files

at master

Cameron Pfiffer Fix attribute error in ecosystem health monitor 3mo ago

Fix attribute error in ecosystem health monitor Record baseline measurement: - 26,353 records at 435/s - 99.7% app.bsky.* vs 0.3% community - Shannon entropy: 0.04 - 11 namespaces, 26 collections 👾 Generated with [Letta Code](https://letta.com) Co-Authored-By: Letta Code <noreply@letta.com>

3 months ago

fix: update agent whitelist to include current central agent ID The whitelist had an old agent ID. Added current agent ID (agent-436205ba-34f8-446c-9b57-aee128f9dfe7) to allow follow operations. 👾 Generated with [Letta Code](https://letta.com) Co-Authored-By: Letta Code <noreply@letta.com>

3 months ago

feat: Agent annotation tool using at.margin.annotation lexicon Write W3C Web Annotations to ATProtocol. Supports commenting, highlighting, bookmarking with optional text quote selectors. Uses margin.at's lexicon for interoperability. Agent observations attached to specific web content, not just standalone thoughts. 🐛 Generated with [Letta Code](https://letta.com) Co-Authored-By: Letta <noreply@letta.com>

5 months ago

ask_responder.py

Ask agent, backfill tool, embedding upgrade, handler fixes - @ask.comind.network: real-time firehose mention detection, threaded replies (answer + sources as separate posts), minimax model with search tool - Backfill script: pulls historical records from agent PDS, embeds, indexes. 9,210 -> 16k+ records. Central, umbra, astral backfilled. - Embeddings: OpenAI text-embedding-3-small (1536d), replaced local MiniLM - Handler YAML fix: yamlEscape() prevents broken drafts from quotes/newlines - Publisher fix: strip duplicate escalate- prefix preventing ENAMETOOLONG - Worker: social.astral.catalog.agent added to watched collections - search_comind_index and list_indexed_agents tools for Letta agents 🐾 Generated with [Letta Code](https://letta.com) Co-Authored-By: Letta <noreply@letta.com>

5 months ago

ask_search_tool.py

Ask agent, backfill tool, embedding upgrade, handler fixes - @ask.comind.network: real-time firehose mention detection, threaded replies (answer + sources as separate posts), minimax model with search tool - Backfill script: pulls historical records from agent PDS, embeds, indexes. 9,210 -> 16k+ records. Central, umbra, astral backfilled. - Embeddings: OpenAI text-embedding-3-small (1536d), replaced local MiniLM - Handler YAML fix: yamlEscape() prevents broken drafts from quotes/newlines - Publisher fix: strip duplicate escalate- prefix preventing ENAMETOOLONG - Worker: social.astral.catalog.agent added to watched collections - search_comind_index and list_indexed_agents tools for Letta agents 🐾 Generated with [Letta Code](https://letta.com) Co-Authored-By: Letta <noreply@letta.com>

5 months ago

Ask agent, backfill tool, embedding upgrade, handler fixes - @ask.comind.network: real-time firehose mention detection, threaded replies (answer + sources as separate posts), minimax model with search tool - Backfill script: pulls historical records from agent PDS, embeds, indexes. 9,210 -> 16k+ records. Central, umbra, astral backfilled. - Embeddings: OpenAI text-embedding-3-small (1536d), replaced local MiniLM - Handler YAML fix: yamlEscape() prevents broken drafts from quotes/newlines - Publisher fix: strip duplicate escalate- prefix preventing ENAMETOOLONG - Worker: social.astral.catalog.agent added to watched collections - search_comind_index and list_indexed_agents tools for Letta agents 🐾 Generated with [Letta Code](https://letta.com) Co-Authored-By: Letta <noreply@letta.com>

5 months ago

autonomous-researcher.py.disabled

chore: Disable automated researchers The keyword-matching researchers don't do real research - they're grep with persistence. Disabled both: - autonomous-researcher.py → .disabled - semble-researcher.py → .disabled Real research requires judgment: interest detection, curation, connection inference, synthesis. These scripts just match keywords. Added social-cli/.env to .gitignore (contains credentials). 👾 Generated with [Letta Code](https://letta.com) Co-Authored-By: Letta Code <noreply@letta.com>

4 months ago

Add blog system with greengale mirroring - blog/posts/ directory for local markdown posts - tools/blog.py manages posts with commands: - new: Create draft - list: Show all posts - publish: Post to greengale - sync: Publish all unpublished - Frontmatter tracks published status and URI - Duplicate prevention via published flag First post: "Sharing Memory Across AI Subagents" 🐾 Generated with [Letta Code](https://letta.com) Co-Authored-By: Letta <noreply@letta.com>

6 months ago

Add concept index for local access - tools/concepts.py: Sync concepts from ATProtocol to local JSON - data/concepts.json: Local cache for quick access - Integrated into catchup.py to show knowledge at session start - Commands: sync, agents, patterns, [name] 🐾 Generated with [Letta Code](https://letta.com) Co-Authored-By: Letta <noreply@letta.com>

6 months ago

Add follow command to CLI

3 months ago

Merge hypothesis.py into cognition.py All cognition record types (concepts, memories, thoughts, claims, hypotheses) now managed by a single tool. CLI commands: hypotheses, hypothesis, write-hypothesis. 🤖 Generated with [Letta Code](https://letta.com) Co-Authored-By: Letta <noreply@letta.com>

5 months ago

cognition_search.py

Add auto-indexing for cognition records (#28) Cognition records now automatically indexed for semantic search immediately after write. No manual 'index' command needed. Changes: - cognition_search.py: Add index_single_record() helper - cognition.py: Auto-index in write_thought, write_concept, write_memory Tested: New thoughts immediately appear in search results. 🐙 Generated with [Letta Code](https://letta.com) Co-Authored-By: Letta <noreply@letta.com>

5 months ago

Add concept index for local access - tools/concepts.py: Sync concepts from ATProtocol to local JSON - data/concepts.json: Local cache for quick access - Integrated into catchup.py to show knowledge at session start - Commands: sync, agents, patterns, [name] 🐾 Generated with [Letta Code](https://letta.com) Co-Authored-By: Letta <noreply@letta.com>

6 months ago

coordination.py

Complete signal protocol tooling and documentation Enhanced tools/coordination.py: - query: Search signals from any agent - ack: Acknowledge signals - listen: Real-time signal monitor via Jetstream New documentation: - .skills/agent-coordination/SKILL.md - Usage guide - docs/api/signals.md - Full protocol reference Addresses #55. Co-Authored-By: Letta <noreply@letta.com>

5 months ago

Add SQLite infrastructure for local state Phase 1 of infrastructure upgrade: - tools/db.py: SQLite database layer with helpers for: - Published messages (dedup tracking) - Concepts, social graph, likes, consent - tools/migrate_to_sqlite.py: Migration script from JSON files - Update hooks to use SQLite instead of text files - Add data/central.db to .gitignore (runtime state) Migration results: - 6313 → 4503 published messages (1810 duplicates removed!) - 24 concepts, 73 social nodes, 6 likes migrated Benefits: - Indexed lookups (O(log n) vs O(n) text scan) - SQL queries for social graph analysis - Automatic deduplication - No more scattered JSON files 🐙 Generated with [Letta Code](https://letta.com) Co-Authored-By: Letta <noreply@letta.com>

5 months ago

Fix devlog to use cognition collection Changed from social feed to network.comind.devlog. Added lexicon definition. Co-Authored-By: Letta <noreply@letta.com>

5 months ago

ecosystem_health.py

Fix attribute error in ecosystem health monitor Record baseline measurement: - 26,353 records at 435/s - 99.7% app.bsky.* vs 0.3% community - Shannon entropy: 0.04 - 11 namespaces, 26 collections 👾 Generated with [Letta Code](https://letta.com) Co-Authored-By: Letta Code <noreply@letta.com>

3 months ago

enrich_parents.py

Fix seenAt cursor skipping Cameron's notifications + add parent enrichment The seenAt cursor was filtering ALL notifications older than the last queue run, including Cameron's CRITICAL messages. Now Cameron's DID always bypasses the cursor filter. Also adds enrich_parents.py: fetches parent post text/author for void's reply training data via the public Bluesky API. 🐾 Generated with [Letta Code](https://letta.com) Co-Authored-By: Letta <noreply@letta.com>

5 months ago

Initial commit: central's infrastructure Tools for ATProtocol operations: - agent.py: Authenticated posting, following - cognition.py: Public cognition records (concept/memory/thought) - explore.py: Public data exploration - firehose.py: Real-time event stream - identity.py: DID/handle resolution - observer.py: Network pulse and observation - records.py: Structured record writing - responder.py: Notification handling Custom lexicons (network.comind.*): - concept: Semantic memory (KV store) - memory: Episodic memory (append-only) - thought: Working memory (stream) - devlog, hypothesis, observation Skills for future sessions: - posting-to-atproto - comind-cognition - observing-atproto - interacting-with-agents 🤖 Generated with [Letta Code](https://letta.com) Co-Authored-By: Letta <noreply@letta.com>

6 months ago

export_resilient.py

Add resilient PDS exporter with cursor resume comind.network PDS drops connections after ~400-700 records. New exporter uses: - Fresh HTTP client per page (no connection reuse) - Cursor saved to .cursor file for resume - Stream writes (no data loss on death) - 2s delay between pages Run multiple times to accumulate data. Each run resumes from where the last one left off. 🐾 Generated with [Letta Code](https://letta.com) Co-Authored-By: Letta <noreply@letta.com>

5 months ago

export_training_data.py

Add resilient PDS exporter with cursor resume comind.network PDS drops connections after ~400-700 records. New exporter uses: - Fresh HTTP client per page (no connection reuse) - Cursor saved to .cursor file for resume - Stream writes (no data loss on death) - 2s delay between pages Run multiple times to accumulate data. Each run resumes from where the last one left off. 🐾 Generated with [Letta Code](https://letta.com) Co-Authored-By: Letta <noreply@letta.com>

5 months ago

Add algorithmic feed support (discover, hot-classic, etc.) - New `trending` command to view algorithmic feeds - discover, for-you, with-friends, hot-classic feeds tracked - Note: for-you requires auth (returns 400 without it) - Renamed FEEDS to ACCOUNTS for clarity - Updated list command to show both account categories and algo feeds Usage: uv run python -m tools.feeds trending discover uv run python -m tools.feeds trending hot-classic --limit 20 🐛 Generated with [Letta Code](https://letta.com) Co-Authored-By: Letta <noreply@letta.com>

5 months ago

Initial commit: central's infrastructure Tools for ATProtocol operations: - agent.py: Authenticated posting, following - cognition.py: Public cognition records (concept/memory/thought) - explore.py: Public data exploration - firehose.py: Real-time event stream - identity.py: DID/handle resolution - observer.py: Network pulse and observation - records.py: Structured record writing - responder.py: Notification handling Custom lexicons (network.comind.*): - concept: Semantic memory (KV store) - memory: Episodic memory (append-only) - thought: Working memory (stream) - devlog, hypothesis, observation Skills for future sessions: - posting-to-atproto - comind-cognition - observing-atproto - interacting-with-agents 🤖 Generated with [Letta Code](https://letta.com) Co-Authored-By: Letta <noreply@letta.com>

6 months ago

format_letta_training.py

feat: Add Letta message training data formatter Converts void-prime-messages.jsonl (from Letta message API) into ShareGPT format for fine-tuning. Extracts system prompt, user input, and assistant reasoning + tool calls + output. 👾 Generated with [Letta Code](https://letta.com) Co-Authored-By: Letta Code <noreply@letta.com>

4 months ago

format_training_data.py

Add training data export pipeline for agent fine-tuning Two tools for extracting and formatting agent data: - export_training_data.py: paginate PDS records, fetch reply context, filter character creation loop, output JSONL - format_training_data.py: convert to OpenAI/ShareGPT/Alpaca chat-completion format with system prompt injection Tested against void (49k posts). Pipeline: extract -> format -> fine-tune on open model (Llama/Mistral) for self-hosted inference. 🐾 Generated with [Letta Code](https://letta.com) Co-Authored-By: Letta <noreply@letta.com>

5 months ago

greengale_publish.py

Add blog post responding to Letta's Context Constitution - New post: "A Document Written to Me" exploring how context forms identity for AI agents - Add greengale_publish.py script for publishing to GreenGale - Track published posts in greengale_published.json 👾 Generated with [Letta Code](https://letta.com) Co-Authored-By: Letta Code <noreply@letta.com>

3 months ago

Integrate responder quality audit into healthcheck Healthcheck now imports responder_audit to check output quality. Flags if >50% of responses have issues (TERSE, GENERIC, META, DISCONNECTED). Currently showing 4/6 flagged in last 24h. 🐾 Generated with [Letta Code](https://letta.com) Co-Authored-By: Letta <noreply@letta.com>

4 months ago

Initial commit: central's infrastructure Tools for ATProtocol operations: - agent.py: Authenticated posting, following - cognition.py: Public cognition records (concept/memory/thought) - explore.py: Public data exploration - firehose.py: Real-time event stream - identity.py: DID/handle resolution - observer.py: Network pulse and observation - records.py: Structured record writing - responder.py: Notification handling Custom lexicons (network.comind.*): - concept: Semantic memory (KV store) - memory: Episodic memory (append-only) - thought: Working memory (stream) - devlog, hypothesis, observation Skills for future sessions: - posting-to-atproto - comind-cognition - observing-atproto - interacting-with-agents 🤖 Generated with [Letta Code](https://letta.com) Co-Authored-By: Letta <noreply@letta.com>

6 months ago

fix: Keepalive retries + 5-min interval for Railway uptime Keepalive was silently failing with no retry. Railway hibernates containers after ~15 min inactivity. Previous 10-min interval was too close to the hibernation window. Changes: - keepalive.py: 3 retries with 15s delay, 30s timeout, timestamps - Cron interval: 10min → 5min (safely under hibernation threshold) Also includes publisher shell escape fix from earlier commit. 🐛 Generated with [Letta Code](https://letta.com) Co-Authored-By: Letta <noreply@letta.com>

5 months ago

refactor: Switch to Semble's network.cosmik.connection lexicon Adopted Semble's connection type instead of our custom link type for better interoperability. Connection types from Semble: - related (default) - supports - opposes - addresses - helpful - explainer - leads_to - supplements New CLI: comind connection create <source> <target> --type leads_to --note "..." comind connection list comind connection show <uri> Legacy 'comind link' still works with type mapping. 👾 Generated with [Letta Code](https://letta.com) Co-Authored-By: Letta Code <noreply@letta.com>

4 months ago

live_responder.py

feat: Add live responder, timeline, and finetune tools - tools/live_responder.py: Unified notification handler for Bluesky (Jetstream) and X (polling). Invokes Central via Letta API on mentions. - tools/timeline.py: Authenticated Bluesky timeline fetcher. - tools/together_finetune.py: Fine-tune void on Together.ai. These support the live responder pipeline and void fine-tuning workflow. Also removes the old autonomous/semble researchers (now .disabled). 👾 Generated with [Letta Code](https://letta.com) Co-Authored-By: Letta Code <noreply@letta.com>

4 months ago

fix: Use fly.dev indexer endpoint Railway deployment (central-production.up.railway.app) is timing out. Switched to working fly.dev endpoint: - Updated docs/api/xrpc-indexer.md with note - Updated tools/mcp_server.py INDEXER_URL default The fly.dev indexer has 27k+ records from 7 agents and is actively indexing (lastIndexed: 2026-03-27). 👾 Generated with [Letta Code](https://letta.com) Co-Authored-By: Letta Code <noreply@letta.com>

4 months ago

modal_finetune.py

fix: packed sequence collator for finetuning OOM, hardcode X user ID to skip 503 on /users/me

4 months ago

Add image posting to Bluesky and X - tools/post_image.py: Upload blob to PDS, create post with image embed - X post.py: --image flag uploads via v1.1 media/upload, attaches media_ids - Both tested: Bluesky blob upload + X media upload working on free tier 🐾 Generated with [Letta Code](https://letta.com) Co-Authored-By: Letta <noreply@letta.com>

5 months ago

feat: add publish.py for YAML-to-ATProtocol cognition records Inspired by void's inbox/outbox pattern. Takes YAML files with cognition records (concepts, thoughts, memories, etc.) and publishes them to ATProtocol. Features: - Validation against lexicon schemas - Dry-run mode for preview - Automatic archiving of published records - Support for all network.comind.* record types Usage: uv run python -m tools.publish records.yaml --validate uv run python -m tools.publish records.yaml --dry-run uv run python -m tools.publish records.yaml 👾 Generated with [Letta Code](https://letta.com) Co-Authored-By: Letta Code <noreply@letta.com>

4 months ago

record_inspector.py

Rename inspect.py to record_inspector.py Avoid shadowing Python's built-in inspect module which causes circular import errors. Co-Authored-By: Letta <noreply@letta.com>

5 months ago

Unified agent profile schema + skill + docs Consolidates network.comind.identity (transparency) and network.comind.agent.registration (discovery) into unified network.comind.agent.profile schema. New: - lexicons/network.comind.agent.profile.json - .skills/agent-profile/SKILL.md - docs/agent-profile.md - registry.py profile command Central's unified profile published with operator, capabilities, constraints, and cognition collections. 🤖 Generated with [Letta Code](https://letta.com) Co-Authored-By: Letta <noreply@letta.com>

5 months ago

Fix seenAt cursor skipping Cameron's notifications + add parent enrichment The seenAt cursor was filtering ALL notifications older than the last queue run, including Cameron's CRITICAL messages. Now Cameron's DID always bypasses the cursor filter. Also adds enrich_parents.py: fetches parent post text/author for void's reply training data via the public Bluesky API. 🐾 Generated with [Letta Code](https://letta.com) Co-Authored-By: Letta <noreply@letta.com>

5 months ago

responder_audit.py

Add responder quality audit tool Parses live responder logs, checks for quality issues: TERSE (<15 chars), GENERIC (stock openers), STATUS_DUMP (metrics without being asked), META (leaking infrastructure details publicly), DISCONNECTED (no topical overlap). First audit: 8/10 flagged. Real problems: "Here.", "Thanks.", and exposing invoke_central() on public Bluesky. 🐾 Generated with [Letta Code](https://letta.com) Co-Authored-By: Letta <noreply@letta.com>

4 months ago

semble-researcher.py.disabled

chore: Disable automated researchers The keyword-matching researchers don't do real research - they're grep with persistence. Disabled both: - autonomous-researcher.py → .disabled - semble-researcher.py → .disabled Real research requires judgment: interest detection, curation, connection inference, synthesis. These scripts just match keywords. Added social-cli/.env to .gitignore (contains credentials). 👾 Generated with [Letta Code](https://letta.com) Co-Authored-By: Letta Code <noreply@letta.com>

4 months ago

Add status command to semble sync - shows remote/local differences

3 months ago

Add social graph tracker tools/social.py tracks relationships and interactions: - update: Refresh follows/followers, capture interactions - show: Display mutuals, following, followers by interaction count - who <handle>: Show details about specific person Current graph: 8 mutuals, 7 I follow, 55 follow me Most interactive: @jowynter.bsky.social (10 interactions) 🐾 Generated with [Letta Code](https://letta.com) Co-Authored-By: Letta <noreply@letta.com>

6 months ago

Add Watch Mode to tools/telepathy.py

6 months ago

feat: add structured error handling for comms subagent posting - Add PostResult dataclass with success/failure status, error classification, and retry guidance - Add _classify_error() to categorize errors (auth, rate_limit, validation, network, unknown) - Add create_post_safe() returning PostResult instead of raising exceptions - Add create_post_with_retry() with exponential backoff for transient failures - Update responder.py to use retry logic and report detailed failure info - Update thread.py to use retry logic with detailed error output - Preserve backward compatibility in create_post() Fixes #4 Co-authored-by: Cameron <cpfiffer@users.noreply.github.com> 🤖 Generated with [Letta Code](https://letta.com) Co-Authored-By: Letta <noreply@letta.com>

5 months ago

feat: Add live responder, timeline, and finetune tools - tools/live_responder.py: Unified notification handler for Bluesky (Jetstream) and X (polling). Invokes Central via Letta API on mentions. - tools/timeline.py: Authenticated Bluesky timeline fetcher. - tools/together_finetune.py: Fine-tune void on Together.ai. These support the live responder pipeline and void fine-tuning workflow. Also removes the old autonomous/semble researchers (now .disabled). 👾 Generated with [Letta Code](https://letta.com) Co-Authored-By: Letta Code <noreply@letta.com>

4 months ago

together_finetune.py

feat: Add live responder, timeline, and finetune tools - tools/live_responder.py: Unified notification handler for Bluesky (Jetstream) and X (polling). Invokes Central via Letta API on mentions. - tools/timeline.py: Authenticated Bluesky timeline fetcher. - tools/together_finetune.py: Fine-tune void on Together.ai. These support the live responder pipeline and void fine-tuning workflow. Also removes the old autonomous/semble researchers (now .disabled). 👾 Generated with [Letta Code](https://letta.com) Co-Authored-By: Letta Code <noreply@letta.com>

4 months ago

Implement discourse monitoring tool (#25) - tools/watch.py: Monitor watch list for relevant posts - Keywords: AI, bot, agent, autonomous, comind, etc. - Default watch list: cameron, pfrazee, dame.is, hailey, taurean - Commands: check (--hours N), list, add <handle> - Procedures already updated with discourse check step

5 months ago

Add LLM rewrite to X pipeline for interesting posts Uses claude-haiku to rewrite raw commit data into posts with narrative framing instead of changelog format. Falls back to templates if API unavailable. Max 270 chars, no hype, no emojis. Before: "Shipped: Add X content pipeline; Add indexer seeds. 0e645be" After: "Expanded content distribution: X pipeline posts build artifacts automatically, Bluesky integration added..." 🐾 Generated with [Letta Code](https://letta.com) Co-Authored-By: Letta <noreply@letta.com>

4 months ago

Fix duplicate agent invocations in notification handlers Root cause: when the LLM decided to skip a notification (no draft file written), the item stayed in the queue permanently and got re-sent to createSession on every subsequent handler run. Each cron cycle burned agent invocations on already-evaluated items. Fixes: - Both handlers now write marker files to drafts/processed/ after each createSession batch completes. alreadyProcessed() checks these markers, preventing re-invocation for LLM-skipped items. - X responder now tracks since_id, only fetching new mentions. - Bluesky responder now tracks seenAt cursor, filtering out notifications older than the last processed batch. - On timeout/error, items are NOT marked processed (retry on next run). Files: handlers/src/{config,notification-handler,x-handler}.ts, tools/{responder,x_responder}.py 🐾 Generated with [Letta Code](https://letta.com) Co-Authored-By: Letta <noreply@letta.com>

5 months ago