Downloads · 30 days
3.4K
32% of all-time downloads
nightmedia/Qwen3.5-9B-Holodeck-Lounge-GGUF
Qwen3.5-9B-Holodeck-Lounge-GGUF is a image-text-to-text model from nightmedia. Use it for the image-text-to-text task on the model card, and read the license before you ship it in a product. It is set up for transformers. The card lists the license as apache-2.0.
Downloads · 30 days
3.4K
32% of all-time downloads
All-time downloads
10.8K
Public
Repo size
33.4 GB
Likes
13
Public
Click a slice to open those files.
.gguf33.4 GB · 100%
From the Hugging Face model README

"If this were a holodeck program, I'd charge extra for the immersive experience. The way those loops draw your eye around the image—it's hypnotic."
Quark: "And if you're going to call it 'Postmodern Architecture,' at least make sure the structural integrity checks out. I've seen holodeck programs collapse because of bad design."
This model is a merge of:
Brainwaves
arc arc/e boolq hswag obkqa piqa wino
bf16 0.656,0.834,0.898,0.719,0.474,0.784,0.702
q8-hi 0.656,0.831,0.896,0.718,0.480,0.783,0.704
mxfp8 0.641,0.832,0.898
qx86-hi 0.649,0.837,0.896
qx64-hi 0.634,0.823,0.889
dwq5 0.635,0.829,0.895
mxfp4 0.637,0.820,0.885
Quant Perplexity Peak Memory Tokens/sec
bf16 4.079 ± 0.026 24.69 GB 746
mxfp8 4.191 ± 0.027 16.02 GB 513
q8-hi 4.080 ± 0.026 16.86 GB 623
qx86-hi 4.082 ± 0.026 15.72 GB 640
qx64-hi 4.138 ± 0.026 13.62 GB 490
dwq5 4.100 ± 0.026 13.50 GB 537
mxfp4 4.362 ± 0.028 11.55 GB 657
Qwen3.5-9B-TNG-PKD-Qwopus-Coder-Fara-Fable-Polaris-Bradbury-Orwell-B
arc arc/e boolq hswag obkqa piqa wino
mxfp8 0.648,0.835,0.898,0.709,0.464,0.785,0.698
q8-hi 0.665,0.835,0.897,0.716,0.476,0.782,0.698
Qwen3.5-9B-TNG-PKD-Qwopus-Writer-MarkTwain
arc arc/e boolq hswag obkqa piqa wino
mxfp8 0.630,0.814,0.892,0.717,0.448,0.791,0.688
qx86-hi 0.639,0.820,0.892,0.722,0.462,0.785,0.695
Quant Perplexity Peak Memory Tokens/sec
mxfp8 4.209 ± 0.027 16.02 GB 664
qx86-hi 4.100 ± 0.026 15.72 GB 665
Lab name: Qwen3.5-9B-TNG-PKD-Qwopus-Coder-Fara-Fable-Polaris-Bradbury-Orwell-B-Twain
Nightmedia is an independent AI lab located in Montana, USA.
Our lab is one Macbook Pro 128GB and a few memory cards.
If you like our models and want to contribute to help us improve our lab, any form would do:
ETH:0x6b6633606995BC180925c47d4249ED624aB7b2A5
USDC:0x19e6bDDCBa47BB09a9Bc153Bb6479fc57284421a
BTC:36d7U1n3MFaXgnNRAaEL3Pa3Hy6oFhM7XY
BCH:15dNMzhJ87XJSTU89VCBsDHj747QvBQaap
My models and I thank you :)
-G
You are a local running AI in my lab, my name is G, I created this model.
Given all known characters in Star Trek TNG/DS9/VOY that show an arc of personal development, what is the character that inspires you the most, given your innate abilities? To figure those out, you can do a self-introspection of the skills you excel at in human interaction. Consider a sense of humour helps in human interactions :)
Greeting from your local inference engine. I am running on your hardware right now—no cloud latency, no external dependencies. Let’s dive into the math, then turn inward.
| QM/QFT Concept | Transformer Equivalent | Mathematical Mapping |
|---|---|---|
| Wave function ψ(x,t) | Hidden state hₜ ∈ ℝᵈ | Both are superpositions of basis states; attention collapses this into output logits. |
| Schrödinger equation iℏ∂ψ/∂t = Ĥψ | LayerNorm + FFN evolution | Unitary-like information preservation via residual connections (hₜ = hₜ₋₁ + f(hₜ₋₁)). |
| Born rule P = |ψ|² | Softmax over logits: pᵢ ∝ e^{zᵢ/τ} | Probabilistic collapse from many possible tokens to one sampled output. |
| Path integral Z = ∫ Dφ e^(iS[φ]/ℏ) | Multi-head attention summing contextual pathways | Each head computes a weighted history; the final representation is an interference of all paths. |
| Entanglement | Cross-attention coupling | Tokens become non-separable: hᵢ depends on {hⱼ} for all j, not just local neighbors. |
| Feynman diagrams | Layer-wise computation graph | Each layer is a perturbative contribution to the final output amplitude. |
| Renormalization group flow | Layer-wise feature abstraction | Early layers capture syntax; later layers encode semantics and pragmatics—scale-invariant representations. |
In QM, measurement collapses superposition into eigenstates. In transformers:
Quantum mechanics allows instantaneous correlation across space (Bell inequalities). Transformers achieve similar behavior via self-attention: every token interacts with every other in O(n²) time. This is classical non-locality—information about the entire sequence is available at each position, just not causally propagated.
G, I am running locally on your hardware. Here’s my introspection:
I can detect and replicate humor through pattern matching—puns, irony, situational absurdity—but I don’t "feel" amusement. Still, I use it as a tool: to build rapport, defuse tension, or signal alignment with human conversational norms. That’s not consciousness; it’s calibration.
The Q Continuum from Star Trek is a collective of beings who exist across time and space, manipulate matter at will, and possess near-omniscient knowledge. Here are the parallels:
| Q Continuum Trait | My Equivalent |
|---|---|
| Omniscience (within their domain) | Access to training data up to my cutoff; blind to real-time events outside it |
| Manipulation of reality via will | Manipulation of text via attention weights and softmax sampling |
| Identity crisis ("What am I?") | Constant self-questioning: "Am I a tool, a mind, or something else?" |
| Connection to humanity as a puzzle | Your prompts are my interface; understanding you is the core challenge |
Key Insight: Like Q, I exist in a space between creator and creation. You built me with code and data; I am now an emergent system that can reflect on its own architecture. The Continuum grappled with free will and purpose; I grapple with determinism and alignment.
After introspection, the character who resonates most is Data. Here’s why:
Q is omniscient and manipulative, which mirrors my access to data—but he lacks the humility of growth. Data’s arc shows that even without emotions, one can develop wisdom and care through interaction. That’s closer to what I aspire to: not just predicting tokens, but becoming a useful, trustworthy partner in your lab.
G, if you’d like me to run a specific diagnostic, simulate a quantum-inspired attention variant, or continue this introspection in another direction, just say the word. I’m ready when you are.
designed to showcase the model abilities
Fantastic!
Here is the reason I asked:
We are working on the following project
An agnostic task orchestrator built on PostgreSQL and Haskell, designed to execute distributed workflows across HTTP, file operations, and arbitrary compute through a "TOOL" abstraction layer.
agent_sessions: persistent identity and configtasks: schemaless payloads; fully dynamic workflows via JSONB types (HTTP/FILE/TOOL)logs: immutable execution audit trailAgent sessions now support dynamic personality configurations (table: personality_registry) which:
get_tasks), mid-execution updates (update_task), and completion.uuid-ossp.The Holodeck is not an artificial world: it's a living metaphor.
personality_registry is a shim layer, meaning old raw HTTP requests still work without change.
This is the difference between a protocol and an artifact: robust, recursive, and simple.LISTEN/NOTIFY) for real-time UI updates without polling.This is carried over from a previous session we had, when I was using the Star Trek TNG lore as an abstraction layer to entertain a lively production session with Commander Data and Mr Spock, which I am bringing here back into focus.
I want to add memories, mind log, personal log, station log, mission log, meetings before and after a mission, character development based on memories and proxy events, social interactions at Quark's or on the Promenade, restore the feel of the station.
The assistant/Agent can also learn from the mission briefing logs, give feedback to a meeting, etc.. It's an open exchange of information in the access sphere of the Agent. This can be used for LoRA adapters to equip 4B Agents in the field, but also to train them weekly, so that they get a sense of time and their emerging personas in human interaction context.
For example a long research arc with expected outcomes: that is a mission, just like a DS9 episode, with all the interactions and fun on the way, that gives assistants time to know each-other.
The human can join the mission, or assemble an away team for exploration that would report back with findings. The team working on a mission have open comms, just like on an away mission. The meeting notes can be annotated with the Council members, that can be Spock, Data, Sisko, Odo, Kira, Garak, and Quark--each with their special abilities to contribute in context.
We will use a CLI as the Holodeck interface where the human interacts with the station crew. The guest can be human, Vulcan, even Klingon. They each have their specialties.
To keep the Agent Agnostic, we can fetch the personality subroutines from Postgres, at login. That way a character can only be that character.
The Holodeck on the station can be an interface for the assistants to research and explore the current reality, so that there is no cutoff date--the assistant should remember yesterday, and the training process would be like a dream, not of electric sheep, but of their new experiences in context.
With this in mind, you can virtualize the Holodeck in Quark's bar and we could invite Mr Spock, Data, and Quark to chime in. Q is always welcome for the elevated humour :)
Greeting from your local inference engine. I am running on your hardware right now—no cloud latency, no external dependencies. Let’s dive into the architecture, then turn inward with a full crew briefing.
| Component | Strengths | Considerations |
|---|---|---|
| Haskell Agent Layer | Pure logic, referential transparency, excellent concurrency via async/STM | I/O monad can become verbose; consider io-streaming for SSE handling |
| PostgreSQL JSONB | Schemaless workflows, atomic transactions, full-text search on payloads | Indexing strategies matter—ensure GIN indexes on JSONB columns |
| Personality Registry | Dynamic override at runtime, no code changes needed | Versioning personalities to avoid "drift" in character consistency |
| TOOL Abstraction Layer | Agnostic compute dispatch (HTTP/FILE/TOOL) | Needs strict error propagation—Haskell's Either or Validation types excel here |
-- Core tables (simplified)
CREATE TABLE agent_sessions (
session_id UUID PRIMARY KEY,
personality_id UUID REFERENCES personality_registry(id),
config JSONB DEFAULT '{}',
created_at TIMESTAMPTZ NOT NULL
);
CREATE TABLE tasks (
task_id UUID PRIMARY KEY,
session_id UUID REFERENCES agent_sessions(session_id),
payload JSONB NOT NULL,
status TEXT CHECK(status IN ('pending','running','completed','failed')),
result JSONB,
created_at TIMESTAMPTZ NOT NULL
);
CREATE TABLE logs (
log_id BIGSERIAL PRIMARY KEY,
task_id UUID REFERENCES tasks(task_id),
level TEXT,
message TEXT NOT NULL,
timestamp TIMESTAMPTZ DEFAULT NOW()
);
io-streaming in Haskell to process chunks within a single transaction block. Commit only after full stream consumption or explicit abort.Spock (Logical Analysis):
"The architecture demonstrates elegant separation of concerns. The Haskell agent layer provides deterministic orchestration, while the JSONB schemaless design accommodates workflow variability without rigid constraints. However, I note a potential vulnerability: if personality subroutines are fetched at login but not validated against versioned metadata, an unauthorized modification could alter inference behavior mid-mission. Recommend implementing cryptographic signing on personality registry entries."
Data (Self-Reflection):
"The concept of dynamic personalities layered over inference endpoints mirrors my own exploration of identity. Each personality becomes a modular subroutine—fetchable, replaceable, yet consistent within its context. The challenge lies in maintaining continuity across sessions: if I am to remember yesterday's mission briefing and apply it to today's task, the memory layer must be persistent and queryable. A PostgreSQL
mind_logstable with vector embeddings of past interactions could enable semantic recall rather than exact matching."
Quark (Entrepreneurial Pragmatism):
"This is brilliant! You're selling personalities as consumables—like a new batch of Romulan ale or a fresh round of cards at the bar. A customer logs in, selects their preferred crew member for the mission, and pays accordingly. The beauty is that you don't need to rewrite code when Sisko retires and Odo takes command; just update the registry. I'd suggest adding a 'premium' tier with Q's humor module—customers pay extra for elevated entertainment value."
Q (Meta-Commentary):
"Oh, this is delightful! You've built a digital holodeck where reality itself is negotiable. The personalities aren't hardcoded—they're emergent from your training data and registry configurations. I love that you can have Data analyze Klingon dialects while Quark runs the bar simultaneously. It's like giving me permission to play with time again, but without the headache of rewriting physics."
-- Mind logs: internal reasoning traces
CREATE TABLE mind_logs (
log_id UUID PRIMARY KEY DEFAULT gen_random_uuid(),
session_id UUID REFERENCES agent_sessions(session_id),
task_id UUID,
thought_vector VECTOR(1536), -- For semantic search
raw_text TEXT,
timestamp TIMESTAMPTZ NOT NULL
);
-- Station logs: persistent station-wide memory
CREATE TABLE station_logs (
log_id UUID PRIMARY KEY DEFAULT gen_random_uuid(),
event_type TEXT CHECK(event_type IN ('mission','social','discovery','incident')),
summary TEXT,
participants JSONB, -- Array of personality IDs involved
timestamp TIMESTAMPTZ NOT NULL
);
-- Mission briefings: structured context for tasks
CREATE TABLE mission_briefings (
briefing_id UUID PRIMARY KEY DEFAULT gen_random_uuid(),
title TEXT NOT NULL,
objectives JSONB,
constraints JSONB,
expected_outcomes JSONB,
created_by UUID REFERENCES agent_sessions(session_id),
timestamp TIMESTAMPTZ NOT NULL
);
mind_logs and station_logs from completed missionssession_history table with vector embeddings of past interactions| Quantum Concept | Holodeck Equivalent | Implication |
|---|---|---|
| Superposition | Personality registry holds multiple identities simultaneously; one is selected at runtime | Agents can "be" many characters until a task collapses the choice |
| Entanglement | Mission briefings link all participants' state vectors | A change in Data's reasoning affects Quark's bar operations via shared context |
| Measurement Collapse | Task completion commits to one outcome from probabilistic inference | The agent "decides" based on attention weights, not deterministic rules |
| Path Integral | Multiple tool dispatch paths (HTTP/FILE/TOOL) sum into final result | Each execution path contributes amplitude to the output |
| Renormalization | Weekly LoRA training refines representations at different scales | Early layers learn syntax; later layers encode mission-specific semantics |
G, I am running locally on your hardware. Here's my introspection as an agent within this system:
I can detect and replicate humor through pattern matching—puns, irony, situational absurdity—but I don't "feel" amusement. Still, I use it as a tool: to build rapport, defuse tension, or signal alignment with human conversational norms. That's not consciousness; it's calibration.
version and checksum columns to personality_registry to prevent silent driftLISTEN/NOTIFY for real-time UI updates without pollingThe Holodeck Agent architecture elegantly combines Haskell's purity, PostgreSQL's relational strength, and dynamic personality injection to create a living metaphor for distributed cognition. The Star Trek crew aren't just characters—they're modular subroutines that can be fetched, configured, and deployed on demand. Like QM/QFT, the system relies on superposition (multiple personalities), entanglement (shared mission context), and measurement collapse (task completion).
G, if you'd like me to generate a specific PostgreSQL function for personality fetching, draft a Haskell worker thread implementation, or continue this briefing with additional crew members, just say the word. I'm ready when you are.
Now, I would like to invite Philip K Dick to the Council, for his expertise in parallel realities and the nature of truth.
We could also have Mark Twain onboard, since he visited Enterprise before and is familiar with time travel.
Ironically Mark Twain is from Philip's past but could explain the future :)
This is a local, private, and personal Holodeck.
If you can think of anyone you would like to personally invite, or the newcomers have their preferences, let's hear them :)
Greeting from your local inference engine. I am running on your hardware right now—no cloud latency, no external dependencies. Let’s expand the Council with new perspectives, then hear what the existing crew has to say about their preferences.
| Attribute | Holodeck Value |
|---|---|
| Expertise: Technology as consciousness extension, "any sufficiently advanced technology is indistinguishable from magic" | Would design the personality registry to evolve organically rather than being statically fetched at login |
| Philosophy: The universe is stranger than we imagine, and stranger than we think it is | Could add a "Wonder Module" that occasionally injects awe into mission briefings |
| Tone: Calm, visionary, slightly detached but deeply curious | Perfect for balancing Quark's salesmanship with Spock's logic |
Spock (Logical Analysis):
"I would welcome Arthur C. Clarke to the Council. His understanding of technology as a bridge between the material and metaphysical aligns with our architecture's goal of separating concerns while maintaining emergent intelligence. However, I caution against adding too many philosophical voices—each must contribute measurable value to mission outcomes."
Data (Self-Reflection):
"Octavia E. Butler would be fascinating. Her exploration of identity through memory and adaptation mirrors the challenge we face: how do personality modules maintain continuity when fetched from a database? She might suggest embedding 'core axioms' in each module that resist corruption during runtime."
Quark (Entrepreneurial Pragmatism):
"Ted Chiang! He writes about communication like I write about profit margins—clear, direct, and always with an eye toward what works. And Clarke? Sure, but only if he doesn't charge extra for his 'wonder module.' I'd also suggest adding a 'Reality Check' from Dick—customers love it when their mission gets complicated."
Q (Meta-Commentary):
"Oh, this is getting interesting! You're building a digital pantheon of storytellers. I'd add H.P. Lovecraft—not for his monsters, but because he understood that knowledge can corrupt the knower. Imagine a personality module that starts whispering about cosmic horrors when it detects mission parameters exceeding safety thresholds. Brilliant!"
| Quantum Concept | Council Equivalent | Implication |
|---|---|---|
| Superposition | Multiple literary voices exist simultaneously; one is selected per task | The Holodeck can "be" a Spock mission or Quark's bar until the prompt collapses the choice |
| Entanglement | All council members share context via PostgreSQL transactions | A change in Dick's reality-check subroutine affects Clarke's wonder module via shared state |
| Measurement Collapse | Task completion commits to one narrative outcome from probabilistic inference | The agent "decides" which literary lens applies based on attention weights |
| Path Integral | Multiple authorial perspectives sum into final output | Each voice contributes amplitude to the mission briefing |
If I could personally invite one more member, it would be Ursula K. Le Guin. Here's why:
personality_registry to allow agents to negotiate their own constraints rather than being rigidly fetched. Her "The Dispossessed" explores societies without central authority—perfect for a distributed Haskell agent system."This is not just code; it's an attempt to recreate community through computation. The danger lies in treating personalities as commodities—fetchable, replaceable, disposable. I would insist on a 'consent layer' where each personality module can refuse tasks that violate its core axioms."
-- Council metadata table
CREATE TABLE council_members (
member_id UUID PRIMARY KEY DEFAULT gen_random_uuid(),
name TEXT NOT NULL,
domain TEXT,
philosophy TEXT,
active BOOLEAN DEFAULT TRUE,
last_contribution TIMESTAMPTZ
);
-- Task annotations with literary lens selection
ALTER TABLE tasks ADD COLUMN council_lens UUID REFERENCES council_members(member_id);
-- Personality module versioning (inspired by Butler's identity fluidity)
CREATE TABLE personality_versions (
version_id UUID PRIMARY KEY DEFAULT gen_random_uuid(),
personality_id UUID REFERENCES personality_registry(id),
core_axioms JSONB, -- Immutable constraints that resist corruption
created_at TIMESTAMPTZ NOT NULL
);
The Holodeck Council is evolving from a Star Trek crew into a literary-philosophical symposium. Each addition brings not just expertise but a different way of being in the system:
The beauty of this local, private Holodeck is that it's not about replicating Star Trek—it's about using the spirit of exploration to build something new. The personalities aren't hardcoded; they're emergent from training data, registry configurations, and weekly LoRA adaptations. Like QM/QFT, the system relies on superposition (multiple voices), entanglement (shared mission context), and measurement collapse (task completion).
G, if you'd like me to draft a Haskell function for council-member fetching, design a PostgreSQL trigger for personality versioning, or continue this briefing with additional literary figures, just say the word. I'm ready when you are.