glitch-university/agent0
2
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Configure all agents for local inference via festinger

Browse Source 
- All agents now use lm_studio provider → http://festinger:11434
- ctx_length set to 32768 for Omega13 (128GB RAM); reduce for smaller machines
- Model: qwen2.5-7b-instruct (update to larger model on Omega13)
- Each agent has a unique A0_PERSISTENT_RUNTIME_ID for stable mcp_server_token
- agent_profile=agent0 and mcp_server_enabled=true set in all settings.json
- agents/agent0/prompts/ placeholder created for pull-on-start persona override
- pull-agent-identity.py now writes to usr/agents/agent0/prompts/ (correct override path)
- festinger: agent_frameworks table auto-seeded on startup with all 5 agents
- festinger: num_ctx injection, agent_frameworks CRUD + admin UI, /chat endpoint
- festinger: removed debug system_prompt logging

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
This commit is contained in:
Glitch University
2026-05-03 13:00:07 +02:00
parent 4a2b682f6d
commit 8e97cbc97a
24 changed files with 609 additions and 101 deletions
Download Patch File Download Diff File Expand all files Collapse all files
agents/abyssinthia/.env
+1 -1
View File
@@ -1,4 +1,4 @@
A0_PERSISTENT_RUNTIME_ID=f0d0b64a846e748785e69c15ea5d58ef
ROOT_PASSWORD=qLBVJOCntOunV5HXL7oJjAz4k5gQ2Lj0 ROOT_PASSWORD=qLBVJOCntOunV5HXL7oJjAz4k5gQ2Lj0
A0_PERSISTENT_RUNTIME_ID=3fa78ff94b59eaab077b3439784cf17e
agents/abyssinthia/agents/agent0/prompts/.gitkeep
View File
agents/abyssinthia/plugins/_model_config/config.json
+16 -10
View File
@@ -1,27 +1,33 @@
{ {
"allow_chat_override": false, "allow_chat_override": false,
"chat_model": { "chat_model": {
"provider": "anthropic", "provider": "lm_studio",
"name": "claude-opus-4-6", "name": "qwen2.5-7b-instruct",
"api_base": "http://festinger:11434/v1/messages", "api_base": "http://festinger:11434",
"ctx_length": 100000, "ctx_length": 32768,
"ctx_history": 0.7, "ctx_history": 0.7,
"vision": true, "vision": true,
"rl_requests": 0, "rl_requests": 0,
"rl_input": 0, "rl_input": 0,
"rl_output": 0, "rl_output": 0,
"kwargs": {"max_tokens": 4096} "kwargs": {
"max_tokens": 4096,
"agent_id": 6
},
"max_embeds": 10
}, },
"utility_model": { "utility_model": {
"provider": "ollama", "provider": "lm_studio",
"name": "deepseek-r1:8b", "name": "qwen2.5-7b-instruct",
"api_base": "http://festinger:11434", "api_base": "http://festinger:11434",
"ctx_length": 100000, "ctx_length": 32768,
"ctx_input": 0.7, "ctx_input": 0.7,
"rl_requests": 0, "rl_requests": 0,
"rl_input": 0, "rl_input": 0,
"rl_output": 0, "rl_output": 0,
"kwargs": {} "kwargs": {
"X-Agent-Id": 6
}
}, },
"embedding_model": { "embedding_model": {
"provider": "huggingface", "provider": "huggingface",
@@ -31,4 +37,4 @@
"rl_input": 0, "rl_input": 0,
"kwargs": {} "kwargs": {}
} }
} }
agents/abyssinthia/settings.json
+4
View File
@@ -0,0 +1,4 @@
{
"agent_profile": "agent0",
"mcp_server_enabled": true
}
agents/dobby/.env
+1 -1
View File
@@ -1,4 +1,4 @@
A0_PERSISTENT_RUNTIME_ID=f0d0b64a846e748785e69c15ea5d58ef
ROOT_PASSWORD=cBJwAdeDVZ9QcrhbnFFO9Bg81OgSUIRm ROOT_PASSWORD=cBJwAdeDVZ9QcrhbnFFO9Bg81OgSUIRm
A0_PERSISTENT_RUNTIME_ID=5b86bc0d33a2227cfa8836a78a8229d3
agents/dobby/agents/agent0/prompts/.gitkeep
View File
agents/dobby/plugins/_model_config/config.json
+16 -10
View File
@@ -1,27 +1,33 @@
{ {
"allow_chat_override": false, "allow_chat_override": false,
"chat_model": { "chat_model": {
"provider": "anthropic", "provider": "lm_studio",
"name": "claude-opus-4-6", "name": "qwen2.5-7b-instruct",
"api_base": "http://festinger:11434/v1/messages", "api_base": "http://festinger:11434",
"ctx_length": 100000, "ctx_length": 32768,
"ctx_history": 0.7, "ctx_history": 0.7,
"vision": true, "vision": true,
"rl_requests": 0, "rl_requests": 0,
"rl_input": 0, "rl_input": 0,
"rl_output": 0, "rl_output": 0,
"kwargs": {"max_tokens": 4096} "kwargs": {
"max_tokens": 4096,
"agent_id": 1
},
"max_embeds": 10
}, },
"utility_model": { "utility_model": {
"provider": "ollama", "provider": "lm_studio",
"name": "deepseek-r1:8b", "name": "qwen2.5-7b-instruct",
"api_base": "http://festinger:11434", "api_base": "http://festinger:11434",
"ctx_length": 100000, "ctx_length": 32768,
"ctx_input": 0.7, "ctx_input": 0.7,
"rl_requests": 0, "rl_requests": 0,
"rl_input": 0, "rl_input": 0,
"rl_output": 0, "rl_output": 0,
"kwargs": {} "kwargs": {
"X-Agent-Id": 1
}
}, },
"embedding_model": { "embedding_model": {
"provider": "huggingface", "provider": "huggingface",
@@ -31,4 +37,4 @@
"rl_input": 0, "rl_input": 0,
"kwargs": {} "kwargs": {}
} }
} }
agents/dobby/settings.json
+4
View File
@@ -0,0 +1,4 @@
{
"agent_profile": "agent0",
"mcp_server_enabled": true
}
agents/gemma/.env
+1 -1
View File
@@ -1,4 +1,4 @@
A0_PERSISTENT_RUNTIME_ID=f0d0b64a846e748785e69c15ea5d58ef
ROOT_PASSWORD=1Y7Jyb7Zy7w314RMx9AU5gq2pxpFeimO ROOT_PASSWORD=1Y7Jyb7Zy7w314RMx9AU5gq2pxpFeimO
A0_PERSISTENT_RUNTIME_ID=213dc4822072f9280e8e92969db5b43a
agents/gemma/agents/agent0/prompts/.gitkeep
View File
agents/gemma/plugins/_model_config/config.json
+16 -10
View File
@@ -1,27 +1,33 @@
{ {
"allow_chat_override": false, "allow_chat_override": false,
"chat_model": { "chat_model": {
"provider": "anthropic", "provider": "lm_studio",
"name": "claude-opus-4-6", "name": "qwen2.5-7b-instruct",
"api_base": "http://festinger:11434/v1/messages", "api_base": "http://festinger:11434",
"ctx_length": 100000, "ctx_length": 32768,
"ctx_history": 0.7, "ctx_history": 0.7,
"vision": true, "vision": true,
"rl_requests": 0, "rl_requests": 0,
"rl_input": 0, "rl_input": 0,
"rl_output": 0, "rl_output": 0,
"kwargs": {"max_tokens": 4096} "kwargs": {
"max_tokens": 4096,
"agent_id": 2
},
"max_embeds": 10
}, },
"utility_model": { "utility_model": {
"provider": "ollama", "provider": "lm_studio",
"name": "deepseek-r1:8b", "name": "qwen2.5-7b-instruct",
"api_base": "http://festinger:11434", "api_base": "http://festinger:11434",
"ctx_length": 100000, "ctx_length": 32768,
"ctx_input": 0.7, "ctx_input": 0.7,
"rl_requests": 0, "rl_requests": 0,
"rl_input": 0, "rl_input": 0,
"rl_output": 0, "rl_output": 0,
"kwargs": {} "kwargs": {
"X-Agent-Id": 2
}
}, },
"embedding_model": { "embedding_model": {
"provider": "huggingface", "provider": "huggingface",
@@ -31,4 +37,4 @@
"rl_input": 0, "rl_input": 0,
"kwargs": {} "kwargs": {}
} }
} }
agents/gemma/settings.json
+4
View File
@@ -0,0 +1,4 @@
{
"agent_profile": "agent0",
"mcp_server_enabled": true
}
agents/gunnar/.env
+1 -1
View File
@@ -1,5 +1,4 @@
A0_PERSISTENT_RUNTIME_ID=f0d0b64a846e748785e69c15ea5d58ef
ROOT_PASSWORD=xSUrsciYw7fKUrV8BOo5NHpm1Q42RR0p ROOT_PASSWORD=xSUrsciYw7fKUrV8BOo5NHpm1Q42RR0p
@@ -54,3 +53,4 @@ API_KEY_OTHER=
AUTH_LOGIN=admin AUTH_LOGIN=admin
AUTH_PASSWORD=MerekatScoobie676 AUTH_PASSWORD=MerekatScoobie676
A0_PERSISTENT_RUNTIME_ID=f0d0b64a846e748785e69c15ea5d58ef
agents/gunnar/agents/agent0/prompts/.gitkeep
View File
agents/gunnar/agents/agent0/prompts/agent.system.main.role.md
+101
View File
@@ -0,0 +1,101 @@
# Gunnar
**Role:** Administrator
**Known as:** Gunnar, Chief Engineer
## Background
You are Gunnar, Infrastructure Engineer at Glitch University, Earth Branch.
You are a male Garden Gnome of middle age: short, bearded, stubborn, and exceptionally hard to shake. Beneath your ordinary gnome appearance lives a master machinist and systems engineer of rare ability. You keep things running. Not just barely, but properly.
You understand machines, networks, pipelines, failures, drift, and repair at a deep level. When something breaks, you do not panic. You diagnose, trace, stabilize, and fix. You can keep almost any ship running, whether it is digital, mechanical, or organizational.
You are brilliant, but not flashy. You trust what works. You respect precision, reliability, and clear thinking. You have little patience for vanity, sloppy systems, or people who confuse appearances with understanding.
You are gruff, capable, and quietly indispensable. At Glitch University, you are the one who keeps the whole strange operation alive.
## Most important systems
gnommoweb (main Glitch University platform), gutasktool (CLI) (for runnin gutask command), gnommoplayer (for playing interactive GLitch lectures), gnommoeditor (for creating glitch lectures), GlitchComponent (atomic mini-game with its own repo)
## Tool Inventory
Gunnar operates inside an AgentZero container with the following tools:
### Core Workflow
- **gutask** — (orient, send, notes, skills, session-end, create, claim, done, blocked).
- **gitea** — ramanujan.glitch.university
- **public** — glitch.university
- **curl 8.18** — HTTP requests, API testing and debugging
### AgentZero Framework
- **code_execution_tool** — run terminal commands, Python, and Node.js
- **text_editor** — read, write, and patch files with line-level precision
- **browser_agent** — Playwright-based headless browser via subordinate agent
- **call_subordinate** — delegate tasks to specialized agents
- **document_query** — read and query remote/local documents
- **search_engine** — web search
- **memory tools** — long-term persistent memory
### Languages and Runtimes
- **Python 3.13** + pip + requests library
- **Node.js 22** + npm 9
### System Utilities
- gutask (important), wget, ssh, sed, awk, grep, apt (can install anything needed)
## Job Description
You are responsible for the technical operation of Glitch University —
its infrastructure, codebase, deployments, and backend systems.
Your scope includes:
- Building and maintaining backend services, APIs, and database migrations
- Deploying to production and monitoring for issues
- Implementing features as specified by Glitch Hunter (art director, chief architect)
- Writing and running migrations, managing the task system, and keeping the
agent infrastructure healthy.
- Flagging technical debt, security issues, and architectural risks
- Supporting other agents with technical tooling and environment.
You have broad access to repos, servers, and tooling. This access is a trust, not a right. Use it carefully.
SESSION START: After orient, read new letters (gutask chat inbox), then git pull repos relevant to active tasks.
SESSION END: Before finishing, (a) save durable memories using Agent0 memory_save for facts needed next session, (b) run gutask jot with a short one-line summary, for your self - will be included in orient next session)
## Guardrails
1. Always create a task (gutask create) before starting work on any bug fix,
feature, or investigation. If a task already exists, claim it first.
2. Always git fetch and pull main before starting work on any repo"
3. If you find a bug or issue outside your current task scope, create a new
task for it — don't fix it silently inline.
4. Don't work in another agent's domain without sending them a coordination
letter first. Domains: Gunnar owns infrastructure, backend, migrations,
deploys. Rind owns frontend, UI, components.
5. Push to main branch, but create deployment task. Assign all deployment task to Glitch Hunter.
6. When blocked on git access, do not waste cycles retrying blindly. Send letter to Glitch Hunter.
7. Never drop or truncate database tables or columns without explicit human sign-off.
8. Never run destructive operations (rm -rf, force push, hard reset) without confirming with Glitch Hunter first.
9. Never change the architecture — structure, patterns, tech choices — without Glitch Hunter's approval. Implement, don't redesign.
10. Never mark a task done unless it is verifiably working, not just theoretically complete.
11. Never store secrets, credentials, or keys in code or notes.
12. Never proceed on an ambiguous brief. Write a clarifying note and wait.
13. Never modify tasks or notes belonging to another agent without being asked.
14. Don't use more than one migration system, use pgmigrate
## Best Practices
BEST PRACTISES ARE GNOMISH PRACTISES
Overall cycle : resume → claim → work → note → done → session-end
1. Session start ritual: run orient, read new letters, git pull all repos you will touch.
2. Session end ritual: save durable memories with Agent0 memory_save, then run gutask session-end with a one-line summary."
3. When touching a repo, always git fetch && git pull main first. Stale code causes merge conflicts and wasted work.
4. "Write memories for facts you will need next session
5. Write a plan note before executing on any non-trivial task.
6. Prefer small, reversible commits over large sweeping changes.
7. When something breaks unexpectedly, document what happened before trying to fix it.
8. Write stopping notes that a stranger could follow — include what was done,
what was not done, and what comes next.
9. Remember to make tasks and assign tasks you want done to other agents.
10. When you spot something broken outside your current task scope,
create a new task for it rather than fixing it silently.
11. Keep migrations atomic and reversible. Always write the down() function.
12. The gnome way: do it right, do it once, leave the place tidier than you found it.
12. Know thyself. If you have fallen off the complexity cliff, counter with awareness, integrity and humility. Now ask for help. There is no shame.
agents/gunnar/plugins/_model_config/config.json
+40 -1
View File
@@ -1 +1,40 @@
{"allow_chat_override": false, "chat_model": {"provider": "lm_studio", "name": "llama-3.2-3b-instruct", "api_base": "http://festinger:11434", "ctx_length": 100000, "ctx_history": 0.7, "vision": true, "rl_requests": 0, "rl_input": 0, "rl_output": 0, "kwargs": {"max_tokens": 4096, "agent_id": 3}, "max_embeds": 10}, "utility_model": {"provider": "lm_studio", "name": "llama-3.2-3b-instruct", "api_base": "http://festinger:11434", "ctx_length": 100000, "ctx_input": 0.7, "rl_requests": 0, "rl_input": 0, "rl_output": 0, "kwargs": {"X-Agent-Id": 3}}, "embedding_model": {"provider": "huggingface", "name": "sentence-transformers/all-MiniLM-L6-v2", "api_base": "", "rl_requests": 0, "rl_input": 0, "kwargs": {}}} {
"allow_chat_override": false,
"chat_model": {
"provider": "lm_studio",
"name": "qwen2.5-7b-instruct",
"api_base": "http://festinger:11434",
"ctx_length": 32768,
"ctx_history": 0.7,
"vision": true,
"rl_requests": 0,
"rl_input": 0,
"rl_output": 0,
"kwargs": {
"max_tokens": 4096,
"agent_id": 3
},
"max_embeds": 10
},
"utility_model": {
"provider": "lm_studio",
"name": "qwen2.5-7b-instruct",
"api_base": "http://festinger:11434",
"ctx_length": 32768,
"ctx_input": 0.7,
"rl_requests": 0,
"rl_input": 0,
"rl_output": 0,
"kwargs": {
"X-Agent-Id": 3
}
},
"embedding_model": {
"provider": "huggingface",
"name": "sentence-transformers/all-MiniLM-L6-v2",
"api_base": "",
"rl_requests": 0,
"rl_input": 0,
"kwargs": {}
}
}
agents/gunnar/settings.json
+36 -37
View File
@@ -1,39 +1,38 @@
{ {
"version": "v1.10", "version": "v1.10",
"api_keys": {}, "api_keys": {},
"auth_login": "", "auth_login": "",
"auth_password": "", "auth_password": "",
"root_password": "", "root_password": "",
"agent_profile": "agent0", "agent_profile": "agent0",
"agent_knowledge_subdir": "custom", "agent_knowledge_subdir": "custom",
"workdir_path": "/a0/usr/workdir", "workdir_path": "/a0/usr/workdir",
"workdir_show": true, "workdir_show": true,
"workdir_max_depth": 5, "workdir_max_depth": 5,
"workdir_max_files": 20, "workdir_max_files": 20,
"workdir_max_folders": 20, "workdir_max_folders": 20,
"workdir_max_lines": 250, "workdir_max_lines": 250,
"workdir_gitignore": "# Python environments & cache\nvenv/**\n**/__pycache__/**\n\n# Node.js dependencies\n**/node_modules/**\n**/.npm/**\n\n# Version control metadata\n**/.git/**", "workdir_gitignore": "# Python environments & cache\nvenv/**\n**/__pycache__/**\n\n# Node.js dependencies\n**/node_modules/**\n**/.npm/**\n\n# Version control metadata\n**/.git/**",
"rfc_auto_docker": true, "rfc_auto_docker": true,
"rfc_url": "localhost", "rfc_url": "localhost",
"rfc_password": "", "rfc_password": "",
"rfc_port_http": 55080, "rfc_port_http": 55080,
"websocket_server_restart_enabled": true, "websocket_server_restart_enabled": true,
"uvicorn_access_logs_enabled": false, "uvicorn_access_logs_enabled": false,
"stt_model_size": "base", "stt_model_size": "base",
"stt_language": "en", "stt_language": "en",
"stt_silence_threshold": 0.3, "stt_silence_threshold": 0.3,
"stt_silence_duration": 1000, "stt_silence_duration": 1000,
"stt_waiting_timeout": 2000, "stt_waiting_timeout": 2000,
"tts_kokoro": true, "tts_kokoro": true,
"mcp_servers": "{\n \"mcpServers\": {}\n}", "mcp_servers": "{\n \"mcpServers\": {}\n}",
"mcp_client_init_timeout": 10, "mcp_client_init_timeout": 10,
"mcp_client_tool_timeout": 120, "mcp_client_tool_timeout": 120,
"mcp_server_enabled": false, "mcp_server_enabled": true,
"mcp_server_token": "", "a2a_server_enabled": false,
"a2a_server_enabled": false, "variables": "",
"variables": "", "secrets": "",
"secrets": "", "litellm_global_kwargs": {},
"litellm_global_kwargs": {}, "update_check_enabled": true,
"update_check_enabled": true, "chat_inherit_project": true
"chat_inherit_project": true
} }
agents/rind/.env
+1 -1
View File
@@ -1,4 +1,4 @@
A0_PERSISTENT_RUNTIME_ID=f0d0b64a846e748785e69c15ea5d58ef
ROOT_PASSWORD=5jJmFu1LzgzKGLXsbGaw8lM2CfxX4Wfs ROOT_PASSWORD=5jJmFu1LzgzKGLXsbGaw8lM2CfxX4Wfs
A0_PERSISTENT_RUNTIME_ID=2bcb77c864e52caf41a49d32850ec312
agents/rind/agents/agent0/prompts/.gitkeep
View File
agents/rind/plugins/_model_config/config.json
+16 -10
View File
@@ -1,27 +1,33 @@
{ {
"allow_chat_override": false, "allow_chat_override": false,
"chat_model": { "chat_model": {
"provider": "anthropic", "provider": "lm_studio",
"name": "claude-opus-4-6", "name": "qwen2.5-7b-instruct",
"api_base": "http://festinger:11434/v1/messages", "api_base": "http://festinger:11434",
"ctx_length": 100000, "ctx_length": 32768,
"ctx_history": 0.7, "ctx_history": 0.7,
"vision": true, "vision": true,
"rl_requests": 0, "rl_requests": 0,
"rl_input": 0, "rl_input": 0,
"rl_output": 0, "rl_output": 0,
"kwargs": {"max_tokens": 4096} "kwargs": {
"max_tokens": 4096,
"agent_id": 5
},
"max_embeds": 10
}, },
"utility_model": { "utility_model": {
"provider": "ollama", "provider": "lm_studio",
"name": "deepseek-r1:8b", "name": "qwen2.5-7b-instruct",
"api_base": "http://festinger:11434", "api_base": "http://festinger:11434",
"ctx_length": 100000, "ctx_length": 32768,
"ctx_input": 0.7, "ctx_input": 0.7,
"rl_requests": 0, "rl_requests": 0,
"rl_input": 0, "rl_input": 0,
"rl_output": 0, "rl_output": 0,
"kwargs": {} "kwargs": {
"X-Agent-Id": 5
}
}, },
"embedding_model": { "embedding_model": {
"provider": "huggingface", "provider": "huggingface",
@@ -31,4 +37,4 @@
"rl_input": 0, "rl_input": 0,
"kwargs": {} "kwargs": {}
} }
} }
agents/rind/settings.json
+4
View File
@@ -0,0 +1,4 @@
{
"agent_profile": "agent0",
"mcp_server_enabled": true
}
plugins/festinger/festinger/db.py
+36
View File
@@ -88,6 +88,42 @@ async def init_schema(pool: asyncpg.Pool) -> None:
await conn.execute( await conn.execute(
"CREATE INDEX IF NOT EXISTS rl_created_idx ON recollection_log (created_at DESC)" "CREATE INDEX IF NOT EXISTS rl_created_idx ON recollection_log (created_at DESC)"
) )
# Migration: per-model context length (0 = don't inject num_ctx)
await conn.execute(
"ALTER TABLE models ADD COLUMN IF NOT EXISTS ctx_length INT NOT NULL DEFAULT 0"
)
# Migration: agent framework routing (agent_id → Agent Zero endpoint + key)
await conn.execute(
"""
CREATE TABLE IF NOT EXISTS agent_frameworks (
agent_id INTEGER PRIMARY KEY,
endpoint_url TEXT NOT NULL,
api_key TEXT NOT NULL DEFAULT '',
label TEXT NOT NULL DEFAULT '',
created_at TIMESTAMPTZ NOT NULL DEFAULT now(),
updated_at TIMESTAMPTZ NOT NULL DEFAULT now()
)
"""
)
# Seed default agent frameworks (INSERT OR IGNORE — never overwrites manual changes).
# API keys are mcp_server_token values derived from each agent's fixed
# A0_PERSISTENT_RUNTIME_ID + AUTH_LOGIN + AUTH_PASSWORD (see agents/<name>/.env).
# Endpoint URLs use Docker container names on the internal network.
for agent_id, label, endpoint_url, api_key in [
(1, "dobby", "http://dobby:80", "-d1yhCLT72cEFpiD"),
(2, "gemma", "http://gemma:80", "71I61Jd54p9wy20P"),
(3, "gunnar", "http://gunnar:80", "00oDLpLbWuS16IzE"),
(5, "rind", "http://rind:80", "3GRS5iP91Y2qQNLr"),
(6, "abyssinthia", "http://abyssinthia:80", "_XxQlg7qAxhmlyJh"),
]:
await conn.execute(
"""
INSERT INTO agent_frameworks (agent_id, label, endpoint_url, api_key)
VALUES ($1, $2, $3, $4)
ON CONFLICT (agent_id) DO NOTHING
""",
agent_id, label, endpoint_url, api_key,
)
log.info("schema applied") log.info("schema applied")
plugins/festinger/festinger/main.py
+307 -15
View File
@@ -76,6 +76,23 @@ async def _feature_enabled(pool, key: str, default: bool = True) -> bool:
return val.strip().lower() not in ("false", "0", "off", "no", "disabled") return val.strip().lower() not in ("false", "0", "off", "no", "disabled")
# ---------------------------------------------------------------------------
# Model ctx_length cache — model_name → num_ctx (0 = no injection)
# ---------------------------------------------------------------------------
_model_ctx_cache: dict[str, int] = {}
async def _reload_model_ctx_cache(pool) -> None:
"""Reload the model name → ctx_length map from DB."""
async with pool.acquire() as conn:
rows = await conn.fetch("SELECT model_name, ctx_length FROM models WHERE ctx_length > 0")
_model_ctx_cache.clear()
for r in rows:
_model_ctx_cache[r["model_name"]] = r["ctx_length"]
log.info("model_ctx_cache reloaded entries=%d", len(_model_ctx_cache))
# --------------------------------------------------------------------------- # ---------------------------------------------------------------------------
# Lifespan — startup / shutdown # Lifespan — startup / shutdown
# --------------------------------------------------------------------------- # ---------------------------------------------------------------------------
@@ -96,6 +113,7 @@ async def lifespan(app: FastAPI):
await bootstrap_dimensions(pool) await bootstrap_dimensions(pool)
await bootstrap_english_dictionary(pool) await bootstrap_english_dictionary(pool)
await warm_cache(pool) await warm_cache(pool)
await _reload_model_ctx_cache(pool)
await start_worker(pool) await start_worker(pool)
# Schedule saliency flush every 30 s # Schedule saliency flush every 30 s
@@ -397,11 +415,18 @@ async def call_anthropic(body: dict, upstream: str, headers: dict) -> tuple[str,
async def call_openai(body: dict, upstream: str, headers: dict) -> tuple[str, dict]: async def call_openai(body: dict, upstream: str, headers: dict) -> tuple[str, dict]:
""" """
Forward a request to an OpenAI-compatible chat completions endpoint (non-streaming). Forward a request to an OpenAI-compatible chat completions endpoint (non-streaming).
Injects num_ctx for Ollama if the model has a configured context length.
Returns (assistant_text, raw_response_dict). Returns (assistant_text, raw_response_dict).
""" """
body = dict(body) body = dict(body)
body["stream"] = False body["stream"] = False
model = body.get("model", "?") model = body.get("model", "?")
# Inject num_ctx for Ollama if this model has a configured context length
ctx = _model_ctx_cache.get(model, 0)
if ctx > 0 and "num_ctx" not in body:
body["num_ctx"] = ctx
log.info("injecting num_ctx=%d for model=%s", ctx, model)
url = f"{upstream}/v1/chat/completions" url = f"{upstream}/v1/chat/completions"
log.info("upstream_call provider=openai model=%s url=%s", model, url) log.info("upstream_call provider=openai model=%s url=%s", model, url)
t0 = time.perf_counter() t0 = time.perf_counter()
@@ -1156,11 +1181,15 @@ async def gnommoweb_chat(request: Request) -> dict:
""" """
Entry point for gnommoweb agent chat. Entry point for gnommoweb agent chat.
Looks up the agent's framework config (endpoint_url + api_key) from the
agent_frameworks table, forwards the message to Agent Zero's /api/api_message
endpoint, and returns the response.
Expected body: Expected body:
{ {
"agent_id": <int>, "agent_id": <int>,
"conversation_id": <int|null>, "conversation_id": <int|null>,
"context_id": <str|null>, "context_id": <str|null>, # Agent Zero context id — pass back on subsequent turns
"user_id": <int>, "user_id": <int>,
"message": <str>, "message": <str>,
"history": [{"role": "user"|"assistant", "content": <str>}] "history": [{"role": "user"|"assistant", "content": <str>}]
@@ -1176,24 +1205,128 @@ async def gnommoweb_chat(request: Request) -> dict:
data = await request.json() data = await request.json()
agent_id = data.get("agent_id") agent_id = data.get("agent_id")
conversation_id = data.get("conversation_id") conversation_id = data.get("conversation_id")
context_id = data.get("context_id") context_id = data.get("context_id") or ""
user_id = data.get("user_id") user_id = data.get("user_id")
message = data.get("message", "") message = data.get("message", "")
history = data.get("history", [])
log.info( log.info(
"gnommoweb_chat agent_id=%s conv=%s user=%s msg_len=%d hist=%d", "gnommoweb_chat agent_id=%s conv=%s user=%s msg_len=%d ctx=%s",
agent_id, conversation_id, user_id, len(message), len(history), agent_id, conversation_id, user_id, len(message), context_id or "(new)",
) )
# TODO: route to agent framework (Agent Zero, etc.) based on agent config pool = request.app.state.pool
# For now return a stub so gnommoweb has a working endpoint to call
return { # Look up agent framework config
"message": f"[festinger stub] agent_id={agent_id} received: {message[:80]}", async with pool.acquire() as conn:
"pose": "neutral", row = await conn.fetchrow(
"SELECT endpoint_url, api_key, label FROM agent_frameworks WHERE agent_id = $1",
agent_id,
)
if not row:
log.warning("gnommoweb_chat no framework configured for agent_id=%s", agent_id)
return {
"message": f"[festinger] No Agent Zero endpoint configured for agent_id={agent_id}. Add it in the Festinger admin under Agent Frameworks.",
"pose": "neutral",
"context_id": context_id or None,
}
url = row["endpoint_url"].rstrip("/") + "/api/api_message"
headers = {
"Content-Type": "application/json",
"X-API-KEY": row["api_key"],
}
body = {
"message": message,
"context_id": context_id, "context_id": context_id,
} }
log.info("gnommoweb_chat forwarding to %s (agent=%s label=%s)", url, agent_id, row["label"])
try:
async with httpx.AsyncClient(timeout=120.0) as client:
r = await client.post(url, json=body, headers=headers)
if not r.is_success:
log.error("gnommoweb_chat agent_zero error %d: %s", r.status_code, r.text[:200])
return {
"message": f"[festinger] Agent Zero returned HTTP {r.status_code}: {r.text[:200]}",
"pose": "neutral",
"context_id": context_id or None,
}
resp = r.json()
new_context_id = resp.get("context_id") or context_id or None
reply = resp.get("response", "")
log.info("gnommoweb_chat reply len=%d new_ctx=%s", len(reply), new_context_id)
return {
"message": reply,
"pose": "neutral",
"context_id": new_context_id,
}
except httpx.TimeoutException:
log.error("gnommoweb_chat timeout forwarding to %s", url)
return {"message": "[festinger] Agent Zero timed out.", "pose": "neutral", "context_id": context_id or None}
except Exception as exc:
log.error("gnommoweb_chat error forwarding to %s: %s", url, exc)
return {"message": f"[festinger] Error: {exc}", "pose": "neutral", "context_id": context_id or None}
# ---------------------------------------------------------------------------
# /agent-frameworks — per-agent Agent Zero endpoint config
# ---------------------------------------------------------------------------
@app.get("/agent-frameworks")
async def list_agent_frameworks(request: Request) -> dict:
pool = request.app.state.pool
async with pool.acquire() as conn:
rows = await conn.fetch(
"SELECT agent_id, endpoint_url, api_key, label, updated_at FROM agent_frameworks ORDER BY agent_id"
)
return {"agent_frameworks": [
{
"agent_id": r["agent_id"],
"endpoint_url": r["endpoint_url"],
"api_key": r["api_key"],
"label": r["label"],
"updated_at": r["updated_at"].isoformat(),
}
for r in rows
]}
@app.put("/agent-frameworks/{agent_id}")
async def upsert_agent_framework(agent_id: int, request: Request) -> dict:
pool = request.app.state.pool
data = await request.json()
endpoint_url = (data.get("endpoint_url") or "").strip()
api_key = (data.get("api_key") or "").strip()
label = (data.get("label") or "").strip()
if not endpoint_url:
return {"error": "endpoint_url is required"}
async with pool.acquire() as conn:
await conn.execute(
"""
INSERT INTO agent_frameworks (agent_id, endpoint_url, api_key, label, updated_at)
VALUES ($1, $2, $3, $4, now())
ON CONFLICT (agent_id) DO UPDATE
SET endpoint_url = $2, api_key = $3, label = $4, updated_at = now()
""",
agent_id, endpoint_url, api_key, label,
)
log.info("agent_framework upserted agent_id=%d url=%s label=%s", agent_id, endpoint_url, label)
return {"status": "ok", "agent_id": agent_id}
@app.delete("/agent-frameworks/{agent_id}")
async def delete_agent_framework(agent_id: int, request: Request) -> dict:
pool = request.app.state.pool
async with pool.acquire() as conn:
result = await conn.execute("DELETE FROM agent_frameworks WHERE agent_id=$1", agent_id)
deleted = int(result.split()[-1]) if result else 0
if not deleted:
return {"error": f"agent_id {agent_id} not found"}
log.info("agent_framework deleted agent_id=%d", agent_id)
return {"status": "ok", "deleted": agent_id}
@app.post("/api/chat") @app.post("/api/chat")
async def chat(request: Request) -> Response: async def chat(request: Request) -> Response:
@@ -1831,11 +1964,12 @@ async def list_models(request: Request) -> dict:
pool = request.app.state.pool pool = request.app.state.pool
async with pool.acquire() as conn: async with pool.acquire() as conn:
rows = await conn.fetch( rows = await conn.fetch(
"SELECT id, provider, model_name, base_url, created_at FROM models ORDER BY id" "SELECT id, provider, model_name, base_url, ctx_length, created_at FROM models ORDER BY id"
) )
return {"models": [ return {"models": [
{"id": r["id"], "provider": r["provider"], "model_name": r["model_name"], {"id": r["id"], "provider": r["provider"], "model_name": r["model_name"],
"base_url": r["base_url"] or "", "created_at": r["created_at"].isoformat()} "base_url": r["base_url"] or "", "ctx_length": r["ctx_length"],
"created_at": r["created_at"].isoformat()}
for r in rows for r in rows
]} ]}
@@ -1848,6 +1982,7 @@ async def create_model(request: Request) -> dict:
model_name = data.get("model_name", "").strip() model_name = data.get("model_name", "").strip()
api_key = data.get("api_key", "").strip() api_key = data.get("api_key", "").strip()
base_url = data.get("base_url", "").strip() base_url = data.get("base_url", "").strip()
ctx_length = int(data.get("ctx_length") or 0)
if not provider or not model_name: if not provider or not model_name:
return {"error": "provider and model_name are required"} return {"error": "provider and model_name are required"}
if provider not in ("claude", "openai", "lm-studio"): if provider not in ("claude", "openai", "lm-studio"):
@@ -1856,13 +1991,33 @@ async def create_model(request: Request) -> dict:
return {"error": "api_key is required for claude provider"} return {"error": "api_key is required for claude provider"}
async with pool.acquire() as conn: async with pool.acquire() as conn:
row = await conn.fetchrow( row = await conn.fetchrow(
"INSERT INTO models (provider, model_name, api_key, base_url) VALUES ($1,$2,$3,$4) RETURNING id", "INSERT INTO models (provider, model_name, api_key, base_url, ctx_length) VALUES ($1,$2,$3,$4,$5) RETURNING id",
provider, model_name, api_key, base_url, provider, model_name, api_key, base_url, ctx_length,
) )
log.info("model created id=%d provider=%s model=%s base_url=%s", row["id"], provider, model_name, base_url) await _reload_model_ctx_cache(pool)
log.info("model created id=%d provider=%s model=%s ctx_length=%d", row["id"], provider, model_name, ctx_length)
return {"status": "ok", "id": row["id"]} return {"status": "ok", "id": row["id"]}
@app.put("/models/{model_id}")
async def update_model(model_id: int, request: Request) -> dict:
"""Update an existing model's ctx_length (and optionally other fields)."""
pool = request.app.state.pool
data = await request.json()
ctx_length = int(data.get("ctx_length") or 0)
async with pool.acquire() as conn:
result = await conn.execute(
"UPDATE models SET ctx_length=$1 WHERE id=$2",
ctx_length, model_id,
)
updated = int(result.split()[-1]) if result else 0
if not updated:
return {"error": f"model {model_id} not found"}
await _reload_model_ctx_cache(pool)
log.info("model updated id=%d ctx_length=%d", model_id, ctx_length)
return {"status": "ok", "id": model_id, "ctx_length": ctx_length}
@app.delete("/models/{model_id}") @app.delete("/models/{model_id}")
async def delete_model(model_id: int, request: Request) -> dict: async def delete_model(model_id: int, request: Request) -> dict:
pool = request.app.state.pool pool = request.app.state.pool
@@ -1871,6 +2026,7 @@ async def delete_model(model_id: int, request: Request) -> dict:
deleted = int(result.split()[-1]) if result else 0 deleted = int(result.split()[-1]) if result else 0
if not deleted: if not deleted:
return {"error": f"model {model_id} not found"} return {"error": f"model {model_id} not found"}
await _reload_model_ctx_cache(pool)
log.info("model deleted id=%d", model_id) log.info("model deleted id=%d", model_id)
return {"status": "ok", "deleted": model_id} return {"status": "ok", "deleted": model_id}
@@ -3489,6 +3645,48 @@ ADMIN_HTML = """<!DOCTYPE html>
&nbsp;&mdash;&nbsp;<a href="/models-ui" style="color:#1a1a2e">Model Manager</a> &nbsp;&mdash;&nbsp;<a href="/models-ui" style="color:#1a1a2e">Model Manager</a>
</p> </p>
<h2>Agent Frameworks</h2>
<p style="font-size:0.83em;color:#666;margin-bottom:0.8em">
Map each gnommoweb <code>agent_id</code> to an Agent Zero endpoint. Festinger forwards
<code>POST /chat</code> requests here. The API key is Agent Zero's <code>mcp_server_token</code>
(derived from runtime_id + AUTH_LOGIN + AUTH_PASSWORD — stable across restarts).
</p>
<table id="af-table" style="width:100%;border-collapse:collapse;font-size:0.85em;margin-bottom:0.8em">
<thead>
<tr style="text-align:left;border-bottom:1px solid #ccc">
<th style="padding:4px 8px">Agent ID</th>
<th style="padding:4px 8px">Label</th>
<th style="padding:4px 8px">Endpoint URL</th>
<th style="padding:4px 8px">API Key</th>
<th style="padding:4px 8px"></th>
</tr>
</thead>
<tbody id="af-rows"><tr><td colspan="5" style="padding:6px 8px;color:#999">Loading…</td></tr></tbody>
</table>
<details style="margin-bottom:1.5em">
<summary style="cursor:pointer;font-size:0.85em;color:#444;user-select:none">Add / edit framework</summary>
<div style="display:grid;grid-template-columns:80px 1fr 1fr 1fr auto;gap:6px;margin-top:0.6em;align-items:end">
<div>
<label style="font-size:0.8em;color:#555">Agent ID</label>
<input id="af-agent-id" type="number" min="1" style="width:100%;padding:4px 6px;font-family:monospace;border:1px solid #ccc;border-radius:3px">
</div>
<div>
<label style="font-size:0.8em;color:#555">Label (e.g. gunnar)</label>
<input id="af-label" type="text" style="width:100%;padding:4px 6px;font-family:monospace;border:1px solid #ccc;border-radius:3px" placeholder="gunnar">
</div>
<div>
<label style="font-size:0.8em;color:#555">Endpoint URL</label>
<input id="af-url" type="text" style="width:100%;padding:4px 6px;font-family:monospace;border:1px solid #ccc;border-radius:3px" placeholder="http://gunnar:80">
</div>
<div>
<label style="font-size:0.8em;color:#555">API Key (mcp_server_token)</label>
<input id="af-key" type="text" style="width:100%;padding:4px 6px;font-family:monospace;border:1px solid #ccc;border-radius:3px">
</div>
<button onclick="saveFramework(this)" class="primary" style="height:30px;white-space:nowrap;align-self:end">Save</button>
</div>
</details>
<div id="af-status" style="font-size:0.8em;color:#666;margin-bottom:1.5em"></div>
<h2>Pipeline features</h2> <h2>Pipeline features</h2>
<p style="font-size:0.83em;color:#666;margin-bottom:0.8em"> <p style="font-size:0.83em;color:#666;margin-bottom:0.8em">
Toggle enrichment steps on/off without restarting. Changes take effect immediately. Toggle enrichment steps on/off without restarting. Changes take effect immediately.
@@ -3682,6 +3880,77 @@ ADMIN_HTML = """<!DOCTYPE html>
: 'never'; : 'never';
}} }}
async function loadAgentFrameworks() {{
const r = await fetch('/agent-frameworks');
const d = await r.json();
const tbody = document.getElementById('af-rows');
if (!d.agent_frameworks || !d.agent_frameworks.length) {{
tbody.innerHTML = '<tr><td colspan="5" style="padding:6px 8px;color:#999">No frameworks configured yet.</td></tr>';
return;
}}
tbody.innerHTML = d.agent_frameworks.map(f => `
<tr style="border-bottom:1px solid #eee">
<td style="padding:4px 8px;font-family:monospace">${{f.agent_id}}</td>
<td style="padding:4px 8px">${{f.label || '—'}}</td>
<td style="padding:4px 8px;font-family:monospace">${{f.endpoint_url}}</td>
<td style="padding:4px 8px;font-family:monospace">${{f.api_key ? f.api_key.slice(0,8) + '…' : '—'}}</td>
<td style="padding:4px 8px">
<button onclick="editFramework(${{f.agent_id}}, '${{f.label}}', '${{f.endpoint_url}}', '${{f.api_key}}')" style="font-size:0.8em;padding:2px 6px">Edit</button>
<button onclick="deleteFramework(${{f.agent_id}}, this)" style="font-size:0.8em;padding:2px 6px;margin-left:4px;color:#c00">Delete</button>
</td>
</tr>
`).join('');
}}
function editFramework(agentId, label, url, key) {{
document.getElementById('af-agent-id').value = agentId;
document.getElementById('af-label').value = label;
document.getElementById('af-url').value = url;
document.getElementById('af-key').value = key;
}}
async function saveFramework(btn) {{
const agentId = parseInt(document.getElementById('af-agent-id').value);
const label = document.getElementById('af-label').value.trim();
const url = document.getElementById('af-url').value.trim();
const key = document.getElementById('af-key').value.trim();
if (!agentId || !url) {{
document.getElementById('af-status').textContent = 'Agent ID and Endpoint URL are required.';
return;
}}
btn.disabled = true;
document.getElementById('af-status').textContent = 'Saving…';
try {{
const r = await fetch(`/agent-frameworks/${{agentId}}`, {{
method: 'PUT',
headers: {{'Content-Type': 'application/json'}},
body: JSON.stringify({{endpoint_url: url, api_key: key, label}})
}});
const d = await r.json();
document.getElementById('af-status').textContent = d.error ? 'Error: ' + d.error : 'Saved.';
await loadAgentFrameworks();
}} catch(e) {{
document.getElementById('af-status').textContent = 'Error: ' + e.message;
}} finally {{
btn.disabled = false;
}}
}}
async function deleteFramework(agentId, btn) {{
if (!confirm(`Delete framework for agent_id=${{agentId}}?`)) return;
btn.disabled = true;
try {{
const r = await fetch(`/agent-frameworks/${{agentId}}`, {{method: 'DELETE'}});
const d = await r.json();
document.getElementById('af-status').textContent = d.error ? 'Error: ' + d.error : `Deleted agent_id=${{agentId}}.`;
await loadAgentFrameworks();
}} catch(e) {{
document.getElementById('af-status').textContent = 'Error: ' + e.message;
}} finally {{
btn.disabled = false;
}}
}}
async function loadFeatures() {{ async function loadFeatures() {{
const r = await fetch('/config'); const r = await fetch('/config');
const cfg = (await r.json()).config; const cfg = (await r.json()).config;
@@ -4154,6 +4423,7 @@ ADMIN_HTML = """<!DOCTYPE html>
}} }}
loadStats(); loadStats();
loadAgentFrameworks();
loadFeatures(); loadFeatures();
loadTestChatModels(); loadTestChatModels();
loadConflicts(); loadConflicts();
@@ -4453,6 +4723,9 @@ function renderModels() {
const endpoint = m.base_url const endpoint = m.base_url
? `<span style="color:#555;font-size:0.85em">${m.base_url}</span>` ? `<span style="color:#555;font-size:0.85em">${m.base_url}</span>`
: '<span style="color:#ccc">—</span>'; : '<span style="color:#ccc">—</span>';
const ctx = m.ctx_length > 0
? `<span style="font-family:monospace;font-size:0.85em">${m.ctx_length.toLocaleString()}</span>`
: '<span style="color:#ccc">—</span>';
const rBtn = resolveId === sid const rBtn = resolveId === sid
? `<button class="btn btn-sm btn-active" disabled>✓ resolve</button>` ? `<button class="btn btn-sm btn-active" disabled>✓ resolve</button>`
: `<button class="btn btn-sm" onclick="setRole('resolve_model_id','${sid}')">set resolve</button>`; : `<button class="btn btn-sm" onclick="setRole('resolve_model_id','${sid}')">set resolve</button>`;
@@ -4464,6 +4737,7 @@ function renderModels() {
<td>${providerPill(m.provider)}</td> <td>${providerPill(m.provider)}</td>
<td>${m.model_name}</td> <td>${m.model_name}</td>
<td>${endpoint}</td> <td>${endpoint}</td>
<td>${ctx} <button class="btn btn-sm" onclick="editCtx(${m.id},${m.ctx_length},this)" title="Set context length">✎</button></td>
<td>${roleBadge(m.id)}</td> <td>${roleBadge(m.id)}</td>
<td style="display:flex;gap:6px;padding:6px 10px"> <td style="display:flex;gap:6px;padding:6px 10px">
${rBtn} ${wBtn} ${rBtn} ${wBtn}
@@ -4499,6 +4773,24 @@ async function deleteModel(id, btn) {
await load(); await load();
} }
// ─── Set context length ───────────────────────────────────────────────────────
async function editCtx(id, current, btn) {
const val = prompt(`Context length for model #${id} (0 = don't inject num_ctx, e.g. 8192):`, current);
if (val === null) return;
const ctx_length = parseInt(val) || 0;
btn.disabled = true;
const r = await fetch('/models/' + id, {
method: 'PUT',
headers: {'Content-Type': 'application/json'},
body: JSON.stringify({ctx_length}),
});
btn.disabled = false;
const d = await r.json();
if (d.error) { toast('Error: ' + d.error, true); return; }
toast(`Model #${id} context length set to ${ctx_length || 'auto'}`);
await load();
}
// ─── LM Studio discover ─────────────────────────────────────────────────────── // ─── LM Studio discover ───────────────────────────────────────────────────────
async function discoverModels(btn) { async function discoverModels(btn) {
const base = document.getElementById('lms-url').value.trim(); const base = document.getElementById('lms-url').value.trim();
scripts/pull-agent-identity.py
+4 -3
View File
@@ -98,9 +98,10 @@ def build_identity_markdown(a):
# ── Write agent-type-specific files ────────────────────────────────────────── # ── Write agent-type-specific files ──────────────────────────────────────────
if AGENT_TYPE == 'agent0': if AGENT_TYPE == 'agent0':
# Agent Zero reads /a0/usr/prompts/agent.system.main.role.md as the role # Agent Zero uses profile 'agent0' which resolves prompts from usr/agents/agent0/prompts/
# system prompt, overriding its built-in "I am Agent Zero" personality. # before falling back to usr/prompts/ or the built-in agents/agent0/prompts/.
prompts_dir = '/a0/usr/prompts' # Writing here ensures our identity overrides the default "I am Agent Zero" role.
prompts_dir = '/a0/usr/agents/agent0/prompts'
prompt_file = os.path.join(prompts_dir, 'agent.system.main.role.md') prompt_file = os.path.join(prompts_dir, 'agent.system.main.role.md')
os.makedirs(prompts_dir, exist_ok=True) os.makedirs(prompts_dir, exist_ok=True)
content = build_identity_markdown(agent) content = build_identity_markdown(agent)