solvi

Build decision systems from a catalog of Python functions and checks plus typed questions, and get answers you can verify.

Why

You describe a task with plain Python functions (computations, checks, answer rules) and questions with typed answers (yes/no, a choice, a score, "not stated", a span of the text, a ranking, a number range). For each request, a strategist plans which functions and checks to run for the asked questions. Every answer comes with:

• a confidence (calibratable per question);
• a reason you can check: the rule inputs, a formula over computed facts, or a quote with character offsets in the source document;
• a hash-chained trace of every step, which can be re-executed later to confirm the answer or pinpoint the step that was altered.

When something cannot be computed, a function fails, or a rule returns an answer outside the allowed options, solvi abstains instead of guessing. A failed hard check always overrides any model confidence.

Grounded decisions. Fuzzy proposes, deterministic decides, everything is in the trace. Each fact and answer records its provenance — given, computed, quoted, decided (a model's choice among options, with probabilities) or learned — and model-backed steps record the model's id and fingerprint, so a replay can tell when the model changed since a decision. A model's quote that is not literally the text at its offsets, or a choice outside its options, is rejected and counted (system.stats); a fallback producer runs or the question abstains. print(res.audit()) shows what each answer rests on, which safeguards fired, and how much of its support is deterministic. A decision without models and one with models are the same system (examples/12_grounded_audit.py).

And it is fast. The strategist plans a flow over a 10 000-part catalog in about 6 ms and runs only the parts the questions need (2.4% of that catalog). Hard checks run first, so a failing one skips the expensive rest; independent slow parts (API calls, model inference) run in parallel. On an insurance-claim desk with slow services (examples/09_strategy_at_scale.py) a full decision takes 463 ms instead of 1 122 ms for a script that computes everything, and 152 ms when an expired policy settles the claim first.

Try it

• solvi playground: write a decision task in Python and run it, watch the strategist's plan, tamper with a trace and see the replay catch it, learn rules from examples.
• solvi documents: cited, typed answers from contracts, invoices, receipts, leases and more — the ModernBERT extractor (ONNX) and the decisions both run in your browser; add a field by describing it.
• solvi arcade: game agents that explain every move — tic-tac-toe, maze, minesweeper, 20 questions, Mafia detective, a bot arena, and "hack the trace".
• solvi realms: an endless strategy game whose factions are solvi systems — tested for 100 000 turns: flat decision time (~0.3–0.6 ms), bounded memory and state, every sampled trace replay OK.
• All run entirely in your browser (Pyodide): no server, no GPU, nothing you type leaves the page.
• Models: solvi-ai/solvi-large (typed decisions, 396M, preview), solvi-ai/solvi-base (the same answers on a CPU / in ONNX, 150M, preview), solvi-ai/extract-base (fields by description) and solvi-ai/extract-receipts. Each model card states what the model was measured on, how well it does, and its limits; all models are listed at huggingface.co/solvi-ai.

Gallery

gallery/ — twelve decision tasks across directions (support triage, email routing, content guard, security alerts, AI-agent audit, release rollout, KYC/AML, clinical screening, credit with adverse-action reasons, procurement 3-way match, double-charge refunds, predictive maintenance), each with scenarios, a runner and a side-by-side against an answer-only model.

Install

pip install solvi              # core: rules, checks, learned answer heads (numpy, scipy, pydantic)
pip install "solvi[model]"     # + torch, transformers: ModernBERT field extractors for documents and the decider
pip install "solvi[onnx]"      # + onnxruntime, tokenizers: the decider (solvi.decide) on CPU without torch
pip install "solvi[serve]"     # + fastapi, uvicorn: `solvi serve app.py:system` — the questions over HTTP (also --mcp)
pip install "solvi[otel]"      # + opentelemetry: decisions as OpenTelemetry spans (solvi.otel)
pip install "solvi[duckdb]"    # + duckdb: stored decisions in a DuckDB file (solvi.DuckDBStorage); [postgres] for PostgreSQL
pip install "solvi[lora]"      # + torch, transformers, peft: part.adapt_lora, a LoRA adapter per question (experimental)

solvi.agents (guarding an agent's tool calls) needs only the core; its adapters use the PydanticAI, LangGraph or OpenAI Agents SDK you already have.

Requires Python 3.10+.

Quickstart (core only, no model)

from datetime import date
from solvi import Answer, Catalog, Question, System

cat = Catalog()

@cat.fn                                            # argument names = facts it reads; function name = fact it sets
def days_requested(start, end):
    return (end - start).days + 1

@cat.fn
def remaining_after(balance, days_requested):
    return balance - days_requested

@cat.check(hard=True, then={"approve": "reject"})  # if this check is False, "approve" is forced to "reject"
def enough_balance(remaining_after):
    return remaining_after >= 0

@cat.check
def enough_notice(start, today, days_requested):
    return days_requested < 5 or (start - today).days >= 14

@cat.rule("approve")
def approve(enough_notice):
    return "approve" if enough_notice else "needs_manager"

system = System(cat, [Question("approve", "Approve the leave?",
                               Answer.choice(["approve", "needs_manager", "reject"]),
                               checkpoints=["enough_balance"])])
res = system.ask({"start": date(2026, 10, 19), "end": date(2026, 10, 23),
                  "today": date(2026, 9, 25), "balance": 14})
print(res["approve"].answer, res["approve"].confidence, res["approve"].why)
print(res.computed_state)
print(res.trace.replay(cat))

Output:

approve 1.0 enough_notice = True
days_requested           = 5
remaining_after          = 9
enough_balance           = True
enough_notice            = True
{'ok': True, 'steps': 5, 'mismatches': [], 'models': [], 'catalog': 'same'}

With "balance": 3 the hard check fails and the answer is reject with status == "forced", whatever the rule says. solvi.show.show(res, cat) prints answers, the planned flow, the computed state and the replay result in one go.

Command line

solvi init triage --with-model && cd triage     # a typed catalog, passing cases.json, README, CI workflow
solvi test . && solvi check catalog.py:system   # regression cases and the catalog lint (what CI runs)
solvi ask catalog.py:system example.json --audit            # one decision and what it rests on (--json, --report html)
solvi models pull solvi-ai/solvi-base           # the only command that downloads; `solvi models` lists, `check` measures
solvi calibrate catalog.py:system route labels.csv --risk 0.1   # act_guard → route.calib.json, loaded by the catalog

Every command is in the guide.

How it works

• Catalog. @cat.fn (computation), @cat.check (bool), @cat.extract (value from text, returned as a Quote with offsets) and @cat.rule(question) (answer rule). A part's contract is its signature: argument names are the facts it reads, the function name is the fact it sets. Type hints are optional and become the facts' types (def risk_score(risk_points: dict[str, float]) -> float): producer and consumer types are checked when a part is registered, values are validated / coerced with pydantic at run time, and a value that fails is rejected like an ungrounded quote (safeguard type_rejected). Untyped parts cost nothing.
• Questions. Question(name, text, Answer.yes_no() | Answer.choice([...]), checkpoints=[...]). Questions without a rule get a small answer head trained from labeled examples (system.fit) or a readable learned rule list (system.learn_rule).
• Strategist. For each question it walks backwards from the rule's arguments (or the learned features) through the catalog signatures to the keys of init_state, adds the question's checkpoints and every check that touches a computed fact. Everything else in the catalog is not executed; the flow records why each part was taken or skipped.
• Execution. Each part runs once, even if several questions need it. Hard checks and their inputs run first; when one fails, the steps only the settled questions needed are skipped (res.trace.skipped). With System(..., workers=8) independent steps run in parallel threads as soon as their inputs are ready. Results go into computed_state with their provenance; extracted values keep their quote. Each step record is hashed and chained to the previous one in flow order, so the trace does not depend on scheduling.
• Answers and trace. res[q].answer / .confidence / .why / .status (ok, forced, abstain), plus res.trace.replay(catalog), which recomputes every step from recorded inputs and reports mismatches, broken hash links and quotes outside the text.

Typed decisions with a model

Types declare questions; the model proposes; checks decide. The fields of a pydantic model are the questions, their types the kinds (one option, several, an ordered score, yes/no); a decider (solvi.decide, solvi[onnx] or solvi[model]) answers them about a text or a JSON / pydantic state — several in one forward pass when the checkpoint can — with probabilities, a calibrated confidence and act / escalate. Hard checks, constraints and rules still decide.

from typing import Literal
from pydantic import BaseModel, Field
from solvi import Catalog, Scale, System
from solvi.decide import DecideModel

class Triage(BaseModel):
    team: Literal["billing", "technical", "shipping"] = Field(description="Which team should handle this ticket?")
    urgency: Scale[Literal["low", "medium", "high", "critical"]] = Field(description="How urgent is it?")
    angry: bool = Field(description="Is the customer angry?")
    topics: list[Literal["refund", "delay", "bug"]] = Field(description="What does the ticket mention?")

model = DecideModel.load("solvi-ai/solvi-base")         # or a local folder; solvi_decide.json says what it can do
cat = Catalog()
questions = model.questions(cat, Triage, text_fact="ticket", escalate_below=0.6)

@cat.check(hard=True, then={"urgency": "critical"})         # a legal threat is critical, whatever the model says
def no_legal_threat(ticket) -> bool:
    return "lawyer" not in str(ticket).lower()
questions[1].checkpoints.append("no_legal_threat")

res = System(cat, questions).ask({"ticket": {"subject": "Charged twice", "body": "Refund my double payment!",
                                             "customer": {"tier": "pro"}}})
print({q: (r.answer, r.status) for q, r in res.results.items()})
print(res.audit("team"))           # probabilities, the model's fingerprint, the shared pass, what escalated and why

A state is read as key paths (customer.tier: pro), the format the decider is trained on; an unsure or escalated answer abstains with the reason (system.stats["model_escalated"], ["low_confidence"]). The checkpoint contract is in docs/decide_format.md; examples/15_typed_decisions.py runs the whole story with a stand-in model. Published deciders are previews: read the model card before relying on one, and fit it on 30–60 labelled examples of your task (part.fit, part.calibrate_for) — checks, constraints and escalation are what make the answers safe to act on, not the model alone.

Every answer is a value and a confidence, and the types also declare answer primitives: Maybe[T] ("not stated" — solvi.Unknown — is a real answer, unlike an abstention), Span[float] (an exact piece of the text, parsed), Rank[...] (the top k, in order), Estimate[0, 7, 14] (a number with an interval), and evidence quotes on any answer (Claim(value, evidence=[...]), Question(require_evidence=True)) — each checked in the text, from rules or a model (guide, examples/16_primitives.py).

Escalation with a guarantee, several models, serving

• A guaranteed risk. part.act_guard(examples, risk=0.10) calibrates on a few hundred labelled examples of your stream so that P(answered alone and wrong) ≤ 10% for inputs like them (conformal risk control); the audit shows the promise behind every answer, or says there is none. part.conformal(examples) gives the person who takes an escalation a short list of candidates. Near ties escalate (min_margin=), and the answer does not depend on the order the options are listed in (sorted by default).
• Any decision model. solvi.systemone.systemone(url, model) puts any POST /v1/systemone service (Jev, Kev, Von, Laya-serve, …) behind your rules, and solvi.llm.llm(base_url, model) any OpenAI-compatible LLM server (OpenAI, OpenRouter, vLLM, llama.cpp, Ollama) — its JSON replies validated, an invalid one escalated, never guessed; Cascade, Vote and Route (solvi.multi) combine models — the next model only when one escalates, an answer only when models of different families agree, or a model picked by code — under one guarantee. On typed-decisions a vote of solvi-large and Julia 1 answered 50% alone against 31% / 40% for each alone, at the same 10% risk (Julia in-distribution there; examples/20_vote_across_families.py).
• Serving and operations. solvi serve module:system exposes the questions over HTTP (OpenAPI from the same types), MCP and the System One API; await system.aask(...) runs async parts concurrently with timeouts; costs="measured" lets the planner pick the fastest equivalent source and switch when it slows down. TraceStorage keeps decisions with a hash chain across them; solvi diff shows which stored decisions a rule or model change would flip; solvi test, solvi check and the honesty suite (solvi honesty) belong in CI; res.report(format="html") and solvi report decisions.db --html out.html give an auditor one page per decision or per period, and solvi.otel.export(res) puts every step in your OpenTelemetry traces. res.counterfactual("approve") says what would have changed the answer ("approve if amount ≤ 1000 (now 1200)"), re-running only the code with the models' recorded proposals held.
• Guarding an agent's tool calls (preview). The agent proposes {"name": tool, "arguments": {...}}; solvi.agents.Guard checks it — the tool is in the catalog, the arguments validate against its types, the values that must come from the conversation are quoted there (and not only from a tool output that says "ignore previous instructions"), your policies (limits, roles, allow-lists) are ordinary hard checks, and an optional decider asks "did the user ask for this?" under act_guard and perturb — then allows it (solvi runs the function), denies it with the reasons, or escalates it to a person. Every decision is a stored, replayable trace. Adapters for PydanticAI, LangGraph and the OpenAI Agents SDK, and solvi serve --guard catalog.py:guard --upstream CMD in front of an MCP server (guide, examples/19_agent_guard.py).
• Text in. system.ask_text("please refund order A-10457, 1.5 million rubles, paid 12 September", decider): the decider picks which question the message asks (or escalates when unsure), each input field is read with a quote and a deterministic parser (numbers, dates, enums, yes / no), missing required fields are listed for a clarifying question, and the trace says those values were read by a model, not given. solvi serve answers texts at POST /ask_text and as the MCP tool ask_text.
• Long documents. decider.decision(..., long="retrieve"): a contract longer than the decider reads is split into sections, BM25 picks the few that bear on the question, the decider reads only those, and quotes point into the whole document; the trace lists the sections read. long="full" reads a text whole up to the length a checkpoint trained on long inputs declares (max_len_long), and retrieves within that length beyond it.
• Learning from corrections. part.memory() escalates an answer when similar corrected cases say another one; fit_fast heads refit on all kept examples as corrections accumulate; part.adapt_lora(examples, holdout=0.3) trains a small LoRA adapter for one question on solvi-base once it has ~100 labelled answers (solvi[lora], experimental); System.learning(store) proposes updates from trusted corrections only and promotes one when it passes held-out, honesty and calibration gates, with rollback (experimental, off unless called) (guide).
• Records you can check later. store.signature() — 64 bytes kept next to the chain's head — later names the one stored record that was edited and restores its hash (preview). solvi.charts draws a chart in which every number is quoted from the text and checked (unit, scale, a pie that adds up), as a deterministic SVG that replays to the same bytes (preview; examples/21_verified_chart.py). The audit, show and the safeguard report render in Russian with System(..., lang="ru").

Planning around dead ends and costs (code strategist)

The default strategist needs the inputs of every producer of a fact. solvi.strategy.ModelStrategist() plans around producers whose inputs are never given, and with producers="equivalent" picks the cheapest verified plan by declared cost= (an exact 0/1 program; hard checks that govern a question always stay in the plan). No model is involved; the plan is one hashed record in the trace and replay re-verifies it.

from solvi.strategy import ModelStrategist
system = System(cat, questions, strategist=ModelStrategist(producers="equivalent"))

A segment model that proposes producers when costs are not declared, and solvi.aliases (wiring parameter names that match no fact), ship as experimental; their weights are not published. See docs/strategist.md and examples/17_model_strategist.py.

Extract from documents

With solvi[model], fields are found by a fine-tuned ModernBERT extractor. The extracted value is always a span of the document, so every answer built on it can be cited.

from solvi.extract_multi import MultiSpanExtractor

ex = MultiSpanExtractor(["total", "date"])          # ModernBERT-large, one pass per document for all fields
ex.fit(train_docs, train_spans, epochs=4)           # train_spans: [{"total": (start, end), "date": (start, end) | None}]
cat.extract(ex.field("total"))                      # doc -> Quote(text, start, end, confidence)
cat.extract(ex.field("date"))

@cat.fn
def amount(total):                                  # extracted values are strings; parse them in ordinary functions
    return float(total.replace(",", ""))

For long documents (contracts), solvi.extract_long.LongSpanExtractor reads the whole text in overlapping 1024-token windows, takes a field description instead of a fixed field list, and supports "no answer" with a per-field threshold. See docs/guide.md.

Results

Same training documents for both sides. The baseline, Laya, is a ModernBERT-large model that answers the typed questions directly, fine-tuned with its authors' recipe. Details and caveats: docs/benchmarks.md.

• Calibrated confidence: ECE 0.008 (SROIE), 0.011 (CORD), 0.027 (CUAD).
• On the document benchmarks, 100% of answers are backed by a quote at stated offsets, or the system abstains.
• Speed: about 0.4 ms per decision when no model is involved; about 39 ms per receipt with the one-pass extractor on an A100 GPU.

Speed

Strategist on random layered catalogs (benchmarks/strategist_scale.py, one CPU core):

Insurance claim desk with six slow services of 100-300 ms (examples/09_strategy_at_scale.py):

A rule-only decision on a small catalog takes well under a millisecond; on documents the extractor dominates (about 39 ms per receipt with the one-pass extractor on an A100).

When to use it

• The answer is a computation or a rule over a few values found in a document or a record: receipts, invoices, contracts, requests, orders.
• You need to show why: auditors, compliance, or a human reviewing low-confidence cases.
• Some rules are non-negotiable (hard checks), and the rest can be learned from about 100 labeled examples.

When not to use it

• Open-ended free-text questions or generated answers. solvi answers typed questions only: yes/no, choices, scores, multi-label, "not stated", exact spans of the text, rankings and number ranges.
• New fields with no labeled examples. Extracting a field from its description alone does not work yet (14% and 66% on two held-out fields); a universal extractor is coming.
• No labels at all. Plan on roughly 100 labeled documents (field positions) per task.
• CPU-only deployment with a quantized model: the int8 ONNX extractor loses up to 12 points on amounts, company names and addresses. fp32 on CPU keeps accuracy but takes about 0.7 s per receipt on 2 cores.

Examples

Run them from a clone: python examples/01_leave_request.py.

More

• docs/guide.md: full API walkthrough.
• docs/decide_format.md: the decider checkpoint contract (for training your own).
• docs/strategist.md: the code strategist and the experimental model strategist and name matching.
• ROADMAP.md, CHANGELOG.md, CONTRIBUTING.md, SECURITY.md.
• docs/benchmarks.md: setups, per-question numbers, caveats.
• benchmarks/: dataset loaders and benchmark scripts (SROIE, CORD, CUAD, Kleister-NDA).
• Tests: pytest.

License

Apache-2.0. See LICENSE.

Citation

A paper is in preparation. Until then, please cite the repository:

@software{solvi,
  title  = {solvi: verifiable decision systems from catalogs of functions and checks},
  author = {mxkuzn and solvi contributors},
  year   = {2026},
  url    = {https://github.com/solvi-ai/solvi}
}